git
12 years agoImprove the description of GIT_PS1_SHOWUPSTREAM
Jonathan "Duke" Leto [Mon, 24 Sep 2012 17:41:26 +0000 (10:41 -0700)] 
Improve the description of GIT_PS1_SHOWUPSTREAM

Describe what '=' means in the output of __git_ps1 when using
GIT_PS1_SHOWUPSTREAM, which was not previously described.

Signed-off-by: Jonathan "Duke" Leto <jonathan@leto.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'nd/maint-diffstat-summary' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:55:31 +0000 (15:55 -0700)] 
Merge branch 'nd/maint-diffstat-summary' into maint

* nd/maint-diffstat-summary:
  Revert diffstat back to English

12 years agoMerge branch 'jw/doc-commit-title' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:55:22 +0000 (15:55 -0700)] 
Merge branch 'jw/doc-commit-title' into maint

* jw/doc-commit-title:
  Documentation: describe subject more precisely

12 years agoMerge branch 'dg/run-command-child-cleanup' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:55:12 +0000 (15:55 -0700)] 
Merge branch 'dg/run-command-child-cleanup' into maint

* dg/run-command-child-cleanup:
  run-command.c: fix broken list iteration in clear_child_for_cleanup

12 years agoMerge branch 'jc/mailinfo-RE' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:55:03 +0000 (15:55 -0700)] 
Merge branch 'jc/mailinfo-RE' into maint

* jc/mailinfo-RE:
  mailinfo: strip "RE: " prefix

12 years agoMerge branch 'sn/ls-remote-get-url-doc' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:54:57 +0000 (15:54 -0700)] 
Merge branch 'sn/ls-remote-get-url-doc' into maint

* sn/ls-remote-get-url-doc:
  ls-remote: document the '--get-url' option

12 years agoMerge branch 'nd/log-n-doc' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:54:43 +0000 (15:54 -0700)] 
Merge branch 'nd/log-n-doc' into maint

* nd/log-n-doc:
  doc: move rev-list option -<n> from git-log.txt to rev-list-options.txt

12 years agoMerge branch 'nd/maint-remote-remove' into maint
Junio C Hamano [Thu, 20 Sep 2012 22:53:31 +0000 (15:53 -0700)] 
Merge branch 'nd/maint-remote-remove' into maint

* nd/maint-remote-remove:
  remote: prefer subcommand name 'remove' to 'rm'

12 years agoDocumentation: Document signature showing options
Stephen Boyd [Thu, 20 Sep 2012 08:10:38 +0000 (01:10 -0700)] 
Documentation: Document signature showing options

The pretty formats for GPG signatures were introduced but never
documented. Use the documentation from the commit that introduced them.
Do the same for the --show-signature option added to git log and
friends.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: fix shell expansion of items
Felipe Contreras [Thu, 20 Sep 2012 02:15:15 +0000 (04:15 +0200)] 
completion: fix shell expansion of items

As reported by Jeroen Meijer[1]; the current code doesn't deal properly
with items (tags, branches, etc.) that have ${} in them because they get
expaned by bash while using compgen.

A simple solution is to quote the items so they get expanded properly
(\$\{\}).

In order to achieve that I took bash-completion's quote() function,
which is rather simple, and renamed it to __git_quote() as per Jeff
King's suggestion.

Solves the original problem for me.

[1] http://article.gmane.org/gmane.comp.version-control.git/201596

Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoGit 1.7.12.1 v1.7.12.1
Junio C Hamano [Tue, 18 Sep 2012 21:34:31 +0000 (14:34 -0700)] 
Git 1.7.12.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'er/doc-fast-import-done' into maint
Junio C Hamano [Tue, 18 Sep 2012 21:33:52 +0000 (14:33 -0700)] 
Merge branch 'er/doc-fast-import-done' into maint

* er/doc-fast-import-done:
  fast-import: document the --done option

12 years agoMerge branch 'jk/config-warn-on-inaccessible-paths' into maint
Junio C Hamano [Tue, 18 Sep 2012 21:24:06 +0000 (14:24 -0700)] 
Merge branch 'jk/config-warn-on-inaccessible-paths' into maint

The attribute system may be asked for a path that itself or its
leading directories no longer exists in the working tree, and it is
fine if we cannot open .gitattribute file in such a case.  Failure
to open per-directory .gitattributes with error status other than
ENOENT and ENOTDIR should be diagnosed.

* jk/config-warn-on-inaccessible-paths:
  attr: failure to open a .gitattributes file is OK with ENOTDIR
  warn_on_inaccessible(): a helper to warn on inaccessible paths
  attr: warn on inaccessible attribute files
  gitignore: report access errors of exclude files
  config: warn on inaccessible files

12 years agoDoc: Improve shallow depth wording
Philip Oakley [Sun, 16 Sep 2012 10:57:58 +0000 (11:57 +0100)] 
Doc: Improve shallow depth wording

Avoid confusion in compound sentence about the start of the commit set
and the depth measure. Use two sentences.

Signed-off-by: Philip Oakley <philipoakley@iee.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation/git-filter-branch: Move note about effect of removing commits
Andreas Schwab [Tue, 18 Sep 2012 15:55:08 +0000 (17:55 +0200)] 
Documentation/git-filter-branch: Move note about effect of removing commits

The note that explains that changes introduced by removed commits are
preserved should be placed directly after the paragraph that describes
such commits removal.  Otherwise the reference to "the commits" appears
out of context.

Also the big example that follows "Consider this history" is about
rewriting part of the history DAG.  Move the paragraph that
describes the operation close to it.

Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot/perf: add "trash directory" to .gitignore
Ramkumar Ramachandra [Mon, 17 Sep 2012 17:06:18 +0000 (22:36 +0530)] 
t/perf: add "trash directory" to .gitignore

Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoAdd missing -z to git check-attr usage text for consistency with man page
Adam Spiers [Mon, 17 Sep 2012 11:38:51 +0000 (12:38 +0100)] 
Add missing -z to git check-attr usage text for consistency with man page

Signed-off-by: Adam Spiers <git@adamspiers.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogit-jump: ignore (custom) prefix in diff mode
Mischa POSLAWSKY [Mon, 17 Sep 2012 01:21:55 +0000 (03:21 +0200)] 
git-jump: ignore (custom) prefix in diff mode

Matching the default file prefix b/ does not yield any results if config
option diff.noprefix or diff.mnemonicprefix is enabled.

Signed-off-by: Mischa POSLAWSKY <git@shiar.nl>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: indent-with-non-tab uses "equivalent tabs" not 8
Wesley J. Landaker [Mon, 17 Sep 2012 14:22:15 +0000 (08:22 -0600)] 
Documentation: indent-with-non-tab uses "equivalent tabs" not 8

Update the documentation of the core.whitespace option
"indent-with-non-tab" to correctly reflect that it catches the use of
spaces instead of the equivalent tabs, rather than a fixed number.

Signed-off-by: Wesley J. Landaker <wjl@icecavern.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocompletion: add --no-edit to git-commit
Yacine Belkadi [Sun, 16 Sep 2012 16:06:11 +0000 (18:06 +0200)] 
completion: add --no-edit to git-commit

Signed-off-by: Yacine Belkadi <yacine.belkadi.1@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDraft release notes to 1.7.12.1
Junio C Hamano [Sat, 15 Sep 2012 04:37:51 +0000 (21:37 -0700)] 
Draft release notes to 1.7.12.1

We are almost there...

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'sb/send-email-reconfirm-fix' into maint
Junio C Hamano [Sat, 15 Sep 2012 04:32:07 +0000 (21:32 -0700)] 
Merge branch 'sb/send-email-reconfirm-fix' into maint

* sb/send-email-reconfirm-fix:
  send-email: initial_to and initial_reply_to are both optional

12 years agoMerge branch 'jc/send-email-reconfirm' into maint
Junio C Hamano [Sat, 15 Sep 2012 04:32:01 +0000 (21:32 -0700)] 
Merge branch 'jc/send-email-reconfirm' into maint

* jc/send-email-reconfirm:
  send-email: validate & reconfirm interactive responses

12 years agoMerge branch 'mz/cherry-pick-cmdline-order' into maint
Junio C Hamano [Sat, 15 Sep 2012 04:24:18 +0000 (21:24 -0700)] 
Merge branch 'mz/cherry-pick-cmdline-order' into maint

* mz/cherry-pick-cmdline-order:
  cherry-pick/revert: respect order of revisions to pick
  demonstrate broken 'git cherry-pick three one two'
  teach log --no-walk=unsorted, which avoids sorting

12 years agoSync with 1.7.11.7
Junio C Hamano [Sat, 15 Sep 2012 04:20:40 +0000 (21:20 -0700)] 
Sync with 1.7.11.7

12 years agoGit 1.7.11.7 v1.7.11.7
Junio C Hamano [Sat, 15 Sep 2012 03:57:23 +0000 (20:57 -0700)] 
Git 1.7.11.7

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jk/maint-quiet-is-synonym-to-s-in-log' into maint-1.7.11
Junio C Hamano [Sat, 15 Sep 2012 03:48:31 +0000 (20:48 -0700)] 
Merge branch 'jk/maint-quiet-is-synonym-to-s-in-log' into maint-1.7.11

* jk/maint-quiet-is-synonym-to-s-in-log:
  log: fix --quiet synonym for -s

12 years agoMerge branch 'jc/maint-ident-missing-human-name' into maint-1.7.11
Junio C Hamano [Sat, 15 Sep 2012 03:48:22 +0000 (20:48 -0700)] 
Merge branch 'jc/maint-ident-missing-human-name' into maint-1.7.11

* jc/maint-ident-missing-human-name:
  split_ident_line(): make best effort when parsing author/committer line

12 years agoMerge branch 'rj/test-regex' into maint-1.7.11
Junio C Hamano [Sat, 15 Sep 2012 03:46:39 +0000 (20:46 -0700)] 
Merge branch 'rj/test-regex' into maint-1.7.11

* rj/test-regex:
  test-regex: Add a test to check for a bug in the regex routines

12 years agoMerge branch 'da/gitk-reload-tag-contents' into maint-1.7.11
Junio C Hamano [Sat, 15 Sep 2012 03:45:55 +0000 (20:45 -0700)] 
Merge branch 'da/gitk-reload-tag-contents' into maint-1.7.11

* da/gitk-reload-tag-contents:
  gitk: Rename 'tagcontents' to 'cached_tagcontent'
  gitk: Teach "Reread references" to reload tags
  gitk: Avoid Meta1-F5

12 years agoMerge branch 'jc/maint-checkout-fileglob-doc' into maint-1.7.11
Junio C Hamano [Sat, 15 Sep 2012 03:45:03 +0000 (20:45 -0700)] 
Merge branch 'jc/maint-checkout-fileglob-doc' into maint-1.7.11

* jc/maint-checkout-fileglob-doc:
  gitcli: contrast wildcard given to shell and to git
  gitcli: formatting fix
  Document file-glob for "git checkout -- '*.c'"

12 years agoRevert diffstat back to English
Nguyễn Thái Ngọc Duy [Thu, 13 Sep 2012 14:16:26 +0000 (21:16 +0700)] 
Revert diffstat back to English

This reverts the i18n part of 7f81463 (Use correct grammar in diffstat
summary line - 2012-02-01) but still keeps the grammar correctness for
English. It also reverts b354f11 (Fix tests under GETTEXT_POISON on
diffstat - 2012-08-27). The result is diffstat always in English
for all commands.

This helps stop users from accidentally sending localized
format-patch'd patches.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoattr: failure to open a .gitattributes file is OK with ENOTDIR
Junio C Hamano [Thu, 13 Sep 2012 19:40:39 +0000 (12:40 -0700)] 
attr: failure to open a .gitattributes file is OK with ENOTDIR

Often we consult an in-tree .gitattributes file that exists per
directory.  Majority of directories do not usually have such a file,
and it is perfectly fine if we cannot open it because there is no
such file, but we do want to know when there is an I/O or permission
error.  Earlier, we made the codepath warn when we fail to open it
for reasons other than ENOENT for that reason.

We however sometimes have to attempt to open the .gitattributes file
from a directory that does not exist in the commit that is currently
checked out.  "git pack-objects" wants to know if a path is marked
with "-delta" attributes, and "git archive" wants to know about
export-ignore and export-subst attributes.  Both commands may and do
need to ask the attributes system about paths in an arbitrary
commit.  "git diff", after removing an entire directory, may want to
know textconv on paths that used to be in that directory.

Make sure we also ignore a failure to open per-directory attributes
file due to ENOTDIR.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocumentation: describe subject more precisely
Jeremy White [Thu, 13 Sep 2012 22:27:09 +0000 (17:27 -0500)] 
Documentation: describe subject more precisely

The discussion of email subject throughout the documentation is
misleading; it indicates that the first line will always become
the subject.  In fact, the subject is generally all lines up until
the first full blank line.

This patch refines that, and makes more use of the concept of a
commit title, with the title being all text up to the first blank line.

Signed-off-by: Jeremy White <jwhite@codeweavers.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoFurther merging in preparation for 1.7.12.1
Junio C Hamano [Wed, 12 Sep 2012 21:10:54 +0000 (14:10 -0700)] 
Further merging in preparation for 1.7.12.1

Describe the following in the draft release notes:

 . jc/apply-binary-p0
 . jc/dotdot-is-parent-directory
 . jc/maint-doc-checkout-b-always-takes-branch-name
 . jk/maint-http-half-auth-push
 . kk/maint-for-each-ref-multi-sort

Yet to be merged before 1.7.12.1 are:

 . jk/config-warn-on-inaccessible-paths
 . jk/maint-quiet-is-synonym-to-s-in-log
 . mz/cherry-pick-cmdline-order

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'maint-1.7.11' into maint
Junio C Hamano [Wed, 12 Sep 2012 21:08:05 +0000 (14:08 -0700)] 
Merge branch 'maint-1.7.11' into maint

12 years agoMerge branch 'jc/apply-binary-p0' into maint-1.7.11
Junio C Hamano [Wed, 12 Sep 2012 21:00:52 +0000 (14:00 -0700)] 
Merge branch 'jc/apply-binary-p0' into maint-1.7.11

"git apply -p0" did not parse pathnames on "diff --git" line
correctly.  This caused patches that had pathnames in no other
places to be mistakenly rejected (most notably, binary patch that
does not rename nor change mode).  Textual patches, renames or mode
changes have preimage and postimage pathnames in different places in
a form that can be parsed unambiguously and did not suffer from this
problem.

* jc/apply-binary-p0:
  apply: compute patch->def_name correctly under -p0

12 years agoMerge branch 'jc/dotdot-is-parent-directory' into maint-1.7.11
Junio C Hamano [Wed, 12 Sep 2012 21:00:34 +0000 (14:00 -0700)] 
Merge branch 'jc/dotdot-is-parent-directory' into maint-1.7.11

"git log .." errored out saying it is both rev range and a path when
there is no disambiguating "--" is on the command line.  Update the
command line parser to interpret ".." as a path in such a case.

* jc/dotdot-is-parent-directory:
  specifying ranges: we did not mean to make ".." an empty set

12 years agoMerge branch 'jc/maint-doc-checkout-b-always-takes-branch-name' into maint-1.7.11
Junio C Hamano [Wed, 12 Sep 2012 20:59:58 +0000 (13:59 -0700)] 
Merge branch 'jc/maint-doc-checkout-b-always-takes-branch-name' into maint-1.7.11

The synopsis said "checkout [-B branch]" to make it clear the
branch name is a parameter to the option, but the heading for the
option description was "-B::", not "-B branch::", making the
documentation misleading.

* jc/maint-doc-checkout-b-always-takes-branch-name:
  doc: "git checkout -b/-B/--orphan" always takes a branch name

12 years agoMerge branch 'jk/maint-http-half-auth-push' into maint-1.7.11
Junio C Hamano [Wed, 12 Sep 2012 20:58:22 +0000 (13:58 -0700)] 
Merge branch 'jk/maint-http-half-auth-push' into maint-1.7.11

Pushing to smart HTTP server with recent Git fails without having
the username in the URL to force authentication, if the server is
configured to allow GET anonymously, while requiring authentication
for POST.

* jk/maint-http-half-auth-push:
  http: prompt for credentials on failed POST
  http: factor out http error code handling
  t: test http access to "half-auth" repositories
  t: test basic smart-http authentication
  t/lib-httpd: recognize */smart/* repos as smart-http
  t/lib-httpd: only route auth/dumb to dumb repos
  t5550: factor out http auth setup
  t5550: put auth-required repo in auth/dumb

12 years agoMerge branch 'kk/maint-for-each-ref-multi-sort' into maint-1.7.11
Junio C Hamano [Wed, 12 Sep 2012 20:57:43 +0000 (13:57 -0700)] 
Merge branch 'kk/maint-for-each-ref-multi-sort' into maint-1.7.11

"git for-each-ref" did not honor multiple "--sort=<key>" arguments
correctly.

* kk/maint-for-each-ref-multi-sort:
  for-each-ref: Fix sort with multiple keys
  t6300: test sort with multiple keys

12 years agoFurther merging down for 1.7.12.1
Junio C Hamano [Tue, 11 Sep 2012 18:27:35 +0000 (11:27 -0700)] 
Further merging down for 1.7.12.1

We will wait for a handful of other fixes that have graduated to the
'master' for 1.8.0 to be tested in the wild and then tag 1.7.12.1:

 . mz/cherry-pick-cmdline-order
 . jk/maint-quiet-is-synonym-to-s-in-log
 . jk/maint-http-half-auth-push
 . jc/apply-binary-p0
 . jk/config-warn-on-inaccessible-paths
 . kk/maint-for-each-ref-multi-sort

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoSync with 1.7.11.6
Junio C Hamano [Tue, 11 Sep 2012 18:23:45 +0000 (11:23 -0700)] 
Sync with 1.7.11.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'js/use-sc-open-max' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:23:06 +0000 (11:23 -0700)] 
Merge branch 'js/use-sc-open-max' into maint

* js/use-sc-open-max:
  sha1_file.c: introduce get_max_fd_limit() helper

12 years agoMerge branch 'js/no-curl-easy-strerror-on-old-curl' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:22:58 +0000 (11:22 -0700)] 
Merge branch 'js/no-curl-easy-strerror-on-old-curl' into maint

* js/no-curl-easy-strerror-on-old-curl:
  http.c: don't use curl_easy_strerror prior to curl-7.12.0

12 years agoGit 1.7.11.6 v1.7.11.6
Junio C Hamano [Tue, 11 Sep 2012 18:18:48 +0000 (11:18 -0700)] 
Git 1.7.11.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'jc/maint-mergetool-style-fix' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:10:23 +0000 (11:10 -0700)] 
Merge branch 'jc/maint-mergetool-style-fix' into maint-1.7.11

* jc/maint-mergetool-style-fix:
  mergetool: style fixes

12 years agoMerge branch 'sz/submodule-force-update' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:10:17 +0000 (11:10 -0700)] 
Merge branch 'sz/submodule-force-update' into maint-1.7.11

* sz/submodule-force-update:
  Make 'git submodule update --force' always check out submodules.

12 years agoMerge branch 'ph/stash-rerere' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:10:12 +0000 (11:10 -0700)] 
Merge branch 'ph/stash-rerere' into maint-1.7.11

* ph/stash-rerere:
  stash: invoke rerere in case of conflict
  test: git-stash conflict sets up rerere

12 years agoMerge branch 'jc/maint-sane-execvp-notdir' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:09:19 +0000 (11:09 -0700)] 
Merge branch 'jc/maint-sane-execvp-notdir' into maint-1.7.11

* jc/maint-sane-execvp-notdir:
  sane_execvp(): ignore non-directory on $PATH

12 years agoMerge branch 'jc/maint-config-exit-status' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:09:09 +0000 (11:09 -0700)] 
Merge branch 'jc/maint-config-exit-status' into maint-1.7.11

* jc/maint-config-exit-status:
  config: "git config baa" should exit with status 1

12 years agoMerge branch 'mh/maint-config-doc-proxy-command' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:09:01 +0000 (11:09 -0700)] 
Merge branch 'mh/maint-config-doc-proxy-command' into maint-1.7.11

* mh/maint-config-doc-proxy-command:
  git-config doc: unconfuse an example
  git-config.txt: fix example

12 years agoMerge branch 'hv/submodule-path-unmatch' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:08:55 +0000 (11:08 -0700)] 
Merge branch 'hv/submodule-path-unmatch' into maint-1.7.11

* hv/submodule-path-unmatch:
  Let submodule command exit with error status if path does not exist

12 years agoMerge branch 'mz/empty-rebase-test' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:08:48 +0000 (11:08 -0700)] 
Merge branch 'mz/empty-rebase-test' into maint-1.7.11

* mz/empty-rebase-test:
  add tests for 'git rebase --keep-empty'

12 years agoMerge branch 'jk/docs-docbook-monospace-display' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:08:40 +0000 (11:08 -0700)] 
Merge branch 'jk/docs-docbook-monospace-display' into maint-1.7.11

* jk/docs-docbook-monospace-display:
  docs: monospace listings in docbook output

12 years agoMerge branch 'ab/diff-write-incomplete-line' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:08:30 +0000 (11:08 -0700)] 
Merge branch 'ab/diff-write-incomplete-line' into maint-1.7.11

* ab/diff-write-incomplete-line:
  Fix '\ No newline...' annotation in rewrite diffs

12 years agoMerge branch 'jc/maint-t7406-rev-parse-max-count-huh' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 18:08:18 +0000 (11:08 -0700)] 
Merge branch 'jc/maint-t7406-rev-parse-max-count-huh' into maint-1.7.11

* jc/maint-t7406-rev-parse-max-count-huh:
  t7406: fix misleading "rev-parse --max-count=1 HEAD"

12 years agoMerge branch 'rr/precompose-utf8-cleanup' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:07:14 +0000 (11:07 -0700)] 
Merge branch 'rr/precompose-utf8-cleanup' into maint

* rr/precompose-utf8-cleanup:
  precompose-utf8: do not call checks for non-ascii "utf8"
  cleanup precompose_utf8

12 years agoMerge branch 'jc/capabilities' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:06:45 +0000 (11:06 -0700)] 
Merge branch 'jc/capabilities' into maint

* jc/capabilities:
  fetch-pack: mention server version with verbose output
  parse_feature_request: make it easier to see feature values
  fetch-pack: do not ask for unadvertised capabilities
  do not send client agent unless server does first
  send-pack: fix capability-sending logic
  include agent identifier in capability string

12 years agoMerge branch 'jc/doc-git-updates' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:06:19 +0000 (11:06 -0700)] 
Merge branch 'jc/doc-git-updates' into maint

* jc/doc-git-updates:
  Documentation: update the introductory section

12 years agoMerge branch 'jk/check-docs-update' into maint
Junio C Hamano [Tue, 11 Sep 2012 18:06:14 +0000 (11:06 -0700)] 
Merge branch 'jk/check-docs-update' into maint

* jk/check-docs-update:
  check-docs: get documented command list from Makefile
  check-docs: drop git-help special-case
  check-docs: list git-gui as a command
  check-docs: factor out command-list
  command-list: mention git-credential-* helpers
  command-list: add git-sh-i18n
  check-docs: update non-command documentation list
  check-docs: mention gitweb specially

12 years agoMerge branch 'tr/void-diff-setup-done' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 17:53:40 +0000 (10:53 -0700)] 
Merge branch 'tr/void-diff-setup-done' into maint-1.7.11

* tr/void-diff-setup-done:
  diff_setup_done(): return void

12 years agoMerge branch 'tr/merge-recursive-flush' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 17:53:31 +0000 (10:53 -0700)] 
Merge branch 'tr/merge-recursive-flush' into maint-1.7.11

* tr/merge-recursive-flush:
  merge-recursive: eliminate flush_buffer() in favor of write_in_full()

12 years agoMerge branch 'nd/index-errno' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 17:53:21 +0000 (10:53 -0700)] 
Merge branch 'nd/index-errno' into maint-1.7.11

* nd/index-errno:
  read_index_from: remove bogus errno assignments

12 years agoMerge branch 'pg/maint-apply-remove-unused-variable' into maint-1.7.11
Junio C Hamano [Tue, 11 Sep 2012 17:53:11 +0000 (10:53 -0700)] 
Merge branch 'pg/maint-apply-remove-unused-variable' into maint-1.7.11

* pg/maint-apply-remove-unused-variable:
  apply: delete unused deflate_origlen from patch struct

12 years agoMerge branch 'jc/test-prereq' into maint
Junio C Hamano [Tue, 11 Sep 2012 17:51:58 +0000 (10:51 -0700)] 
Merge branch 'jc/test-prereq' into maint

* jc/test-prereq:
  t3910: use the UTF8_NFD_TO_NFC test prereq
  test-lib: provide UTF8 behaviour as a prerequisite
  t0050: use the SYMLINKS test prereq
  t0050: use the CASE_INSENSITIVE_FS test prereq
  test-lib: provide case insensitivity as a prerequisite
  test: allow prerequisite to be evaluated lazily
  test: rename $satisfied to $satisfied_prereq

12 years agorun-command.c: fix broken list iteration in clear_child_for_cleanup
David Gould [Tue, 11 Sep 2012 14:32:47 +0000 (15:32 +0100)] 
run-command.c: fix broken list iteration in clear_child_for_cleanup

Iterate through children_to_clean using 'next' fields but with an
extra level of indirection. This allows us to update the chain when
we remove a child and saves us managing several variables around
the loop mechanism.

Signed-off-by: David Gould <david@optimisefitness.com>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoStart preparing for 1.7.12.1
Junio C Hamano [Mon, 10 Sep 2012 22:39:15 +0000 (15:39 -0700)] 
Start preparing for 1.7.12.1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'bc/prune-info' into maint
Junio C Hamano [Mon, 10 Sep 2012 22:37:37 +0000 (15:37 -0700)] 
Merge branch 'bc/prune-info' into maint

"git prune" without "-v" used to warn about leftover temporary
files (which is an indication of an earlier aborted operation).

* bc/prune-info:
  prune.c: only print informational message in show_only or verbose mode

12 years agoMerge branch 'maint-1.7.11' into maint
Junio C Hamano [Mon, 10 Sep 2012 22:31:06 +0000 (15:31 -0700)] 
Merge branch 'maint-1.7.11' into maint

* maint-1.7.11:
  Almost 1.7.11.6
  gitweb: URL-decode $my_url/$my_uri when stripping PATH_INFO
  rebase -i: use full onto sha1 in reflog
  sh-setup: protect from exported IFS
  receive-pack: do not leak output from auto-gc to standard output
  t/t5400: demonstrate breakage caused by informational message from prune
  setup: clarify error messages for file/revisions ambiguity
  send-email: improve RFC2047 quote parsing
  fsck: detect null sha1 in tree entries
  do not write null sha1s to on-disk index
  diff: do not use null sha1 as a sentinel value

12 years agoAlmost 1.7.11.6
Junio C Hamano [Mon, 10 Sep 2012 22:30:46 +0000 (15:30 -0700)] 
Almost 1.7.11.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoMerge branch 'mg/rebase-i-onto-reflog-in-full' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:26:03 +0000 (15:26 -0700)] 
Merge branch 'mg/rebase-i-onto-reflog-in-full' into maint-1.7.11

The reflog entries left by "git rebase" and "git rebase -i" were
inconsistent (the interactive one gave an abbreviated object name).

* mg/rebase-i-onto-reflog-in-full:
  rebase -i: use full onto sha1 in reflog

12 years agoMerge branch 'jc/maint-protect-sh-from-ifs' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:25:45 +0000 (15:25 -0700)] 
Merge branch 'jc/maint-protect-sh-from-ifs' into maint-1.7.11

When the user exports a non-default IFS without HT, scripts that
rely on being able to parse "ls-files -s | while read a b c..."
start to fail.  Protect them from such a misconfiguration.

* jc/maint-protect-sh-from-ifs:
  sh-setup: protect from exported IFS

12 years agoMerge branch 'bc/receive-pack-stdout-protection' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:25:09 +0000 (15:25 -0700)] 
Merge branch 'bc/receive-pack-stdout-protection' into maint-1.7.11

When "git push" triggered the automatic gc on the receiving end, a
message from "git prune" that said it was removing cruft leaked to
the standard output, breaking the communication protocol.

* bc/receive-pack-stdout-protection:
  receive-pack: do not leak output from auto-gc to standard output
  t/t5400: demonstrate breakage caused by informational message from prune

12 years agoMerge branch 'jk/maint-null-in-trees' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:24:53 +0000 (15:24 -0700)] 
Merge branch 'jk/maint-null-in-trees' into maint-1.7.11

"git diff" had a confusion between taking data from a path in the
working tree and taking data from an object that happens to have
name 0{40} recorded in a tree.

* jk/maint-null-in-trees:
  fsck: detect null sha1 in tree entries
  do not write null sha1s to on-disk index
  diff: do not use null sha1 as a sentinel value

12 years agoMerge branch 'tr/maint-send-email-2047' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:24:40 +0000 (15:24 -0700)] 
Merge branch 'tr/maint-send-email-2047' into maint-1.7.11

"git send-email" did not unquote encoded words that appear on the
header correctly, and lost "_" from strings.

* tr/maint-send-email-2047:
  send-email: improve RFC2047 quote parsing

12 years agoMerge branch 'mm/die-with-dashdash-help' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:24:21 +0000 (15:24 -0700)] 
Merge branch 'mm/die-with-dashdash-help' into maint-1.7.11

When the user gives an argument that can be taken as both a
revision name and a pathname without disambiguating with "--", we
used to give a help message "Use '--' to separate".  The message
has been clarified to show where that '--' goes on the command
line.

* mm/die-with-dashdash-help:
  setup: clarify error messages for file/revisions ambiguity

12 years agoMerge branch 'js/gitweb-path-info-unquote' into maint-1.7.11
Junio C Hamano [Mon, 10 Sep 2012 22:23:46 +0000 (15:23 -0700)] 
Merge branch 'js/gitweb-path-info-unquote' into maint-1.7.11

"gitweb" when used with PATH_INFO failed to notice directories with
SP (and other characters that need URL-style quoting) in them.

* js/gitweb-path-info-unquote:
  gitweb: URL-decode $my_url/$my_uri when stripping PATH_INFO

12 years agogitcli: contrast wildcard given to shell and to git
Junio C Hamano [Fri, 7 Sep 2012 20:49:15 +0000 (13:49 -0700)] 
gitcli: contrast wildcard given to shell and to git

People who are not used to working with shell may intellectually
understand how the command line argument is massaged by the shell
but still have a hard time visualizing the difference between
letting the shell expand fileglobs and having Git see the fileglob
to use as a pathspec.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitcli: formatting fix
Junio C Hamano [Mon, 10 Sep 2012 19:47:38 +0000 (12:47 -0700)] 
gitcli: formatting fix

The paragraph to encourage use of "--" in scripts belongs to the
bullet point that describes the behaviour for a command line without
the explicit "--" disambiguation; it is not a supporting explanation
for the entire bulletted list, and it is wrong to make it a separate
paragraph outside the list.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agomailinfo: strip "RE: " prefix
Junio C Hamano [Sun, 9 Sep 2012 09:59:58 +0000 (02:59 -0700)] 
mailinfo: strip "RE: " prefix

We already strip the more common Re: and re:, and we do not often
see RE: from saner MUA, but this prefix does exist and gets used
from time to time.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitk: Rename 'tagcontents' to 'cached_tagcontent'
David Aguilar [Sat, 8 Sep 2012 19:53:16 +0000 (12:53 -0700)] 
gitk: Rename 'tagcontents' to 'cached_tagcontent'

Name the 'tagcontents' variable similarly to the rest of the
variables cleared in the changedrefs() function.

This makes the naming consistent and provides a hint that it
should be cleared when reloading gitk's cache.

Suggested-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agogitk: Teach "Reread references" to reload tags
David Aguilar [Sat, 8 Sep 2012 19:03:13 +0000 (12:03 -0700)] 
gitk: Teach "Reread references" to reload tags

Tag contents, once read, are forever cached in memory.
This makes gitk unable to notice when tag contents change.

Allow users to cause a reload of the tag contents by using
the "File->Reread references" action.

Reported-by: Tim McCormack <cortex@brainonfire.net>
Suggested-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agols-remote: document the '--get-url' option
Stefan Naewe [Fri, 7 Sep 2012 06:41:19 +0000 (08:41 +0200)] 
ls-remote: document the '--get-url' option

While looking for a way to expand the URL of a remote
that uses a 'url.<name>.insteadOf' config option I stumbled
over the undocumented '--get-url' option of 'git ls-remote'.
This adds some minimum documentation for that option.

And while at it, also add that option to the '-h' output.

Signed-off-by: Stefan Naewe <stefan.naewe@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agosend-email: initial_to and initial_reply_to are both optional
Stephen Boyd [Thu, 6 Sep 2012 18:31:11 +0000 (11:31 -0700)] 
send-email: initial_to and initial_reply_to are both optional

We may pick up additional recipients from the format-patch output
files we are sending, in which case it is perfectly valid to leave
the @initial_to empty when the prompt asks.  We may want to start
a new discussion thread without replying to anything, and it is
valid to leave $initial_reply_to empty.

An earlier update to avoid y@example.com stuffed in address fields
did not take these two cases into account.

Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agodoc: move rev-list option -<n> from git-log.txt to rev-list-options.txt
Nguyễn Thái Ngọc Duy [Thu, 6 Sep 2012 14:28:03 +0000 (21:28 +0700)] 
doc: move rev-list option -<n> from git-log.txt to rev-list-options.txt

rev-list-options.txt is included in git-rev-list.txt. This makes sure
rev-list man page also shows that, and at one place, together with
equivalent options -n and --max-count.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoremote: prefer subcommand name 'remove' to 'rm'
Nguyễn Thái Ngọc Duy [Thu, 6 Sep 2012 12:25:23 +0000 (19:25 +0700)] 
remote: prefer subcommand name 'remove' to 'rm'

All remote subcommands are spelled out words except 'rm'. 'rm', being a
popular UNIX command name, may mislead users that there are also 'ls' or
'mv'. Use 'remove' to fit with the rest of subcommands.

'rm' is still supported and used in the test suite. It's just not
widely advertised.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoDocument file-glob for "git checkout -- '*.c'"
Junio C Hamano [Tue, 4 Sep 2012 15:28:27 +0000 (08:28 -0700)] 
Document file-glob for "git checkout -- '*.c'"

Just like we give a similar example in "git add" documentation.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agotest-regex: Add a test to check for a bug in the regex routines
Ramsay Jones [Sat, 1 Sep 2012 17:46:54 +0000 (18:46 +0100)] 
test-regex: Add a test to check for a bug in the regex routines

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agosplit_ident_line(): make best effort when parsing author/committer line
Junio C Hamano [Fri, 31 Aug 2012 21:54:18 +0000 (14:54 -0700)] 
split_ident_line(): make best effort when parsing author/committer line

Commits made by ancient version of Git allowed committer without
human readable name, like this (00213b17c in the kernel history):

    tree 6947dba41f8b0e7fe7bccd41a4840d6de6a27079
    parent 352dd1df32e672be4cff71132eb9c06a257872fe
    author Petr Baudis <pasky@ucw.cz> 1135223044 +0100
    committer  <sam@mars.ravnborg.org> 1136151043 +0100

    kconfig: Remove support for lxdialog --checklist

    ...

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
When fed such a commit, --format='%ci' fails to parse it, and gives
back an empty string.  Update the split_ident_line() to be a bit
more lenient when parsing, but make sure the caller that wants to
pick up sane value from its return value does its own validation.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agocherry-pick/revert: respect order of revisions to pick
Martin von Zweigbergk [Wed, 29 Aug 2012 06:15:56 +0000 (23:15 -0700)] 
cherry-pick/revert: respect order of revisions to pick

When giving multiple individual revisions to cherry-pick or revert, as
in 'git cherry-pick A B' or 'git revert B A', one would expect them to
be picked/reverted in the order given on the command line. They are
instead ordered by their commit timestamp -- in chronological order
for "cherry-pick" and in reverse chronological order for
"revert". This matches the order in which one would usually give them
on the command line, making this bug somewhat hard to notice. Still,
it has been reported at least once before [1].

It seems like the chronological sorting happened by accident because
the revision walker has traditionally always sorted commits in reverse
chronological order when rev_info.no_walk was enabled. In the case of
'git revert B A' where B is newer than A, this sorting is a no-op. For
'git cherry-pick A B', the sorting would reverse the arguments, but
because the sequencer also flips the rev_info.reverse flag when
picking (as opposed to reverting), the end result is a chronological
order. The rev_info.reverse flag was probably flipped so that the
revision walker emits B before C in 'git cherry-pick A..C'; that it
happened to effectively undo the unexpected sorting done when not
walking, was probably a coincidence that allowed this bug to happen at
all.

Fix the bug by telling the revision walker not to sort the commits
when not walking. The only case we want to reverse the order is now
when cherry-picking and walking revisions (rev_info.no_walk = 0).

 [1] http://thread.gmane.org/gmane.comp.version-control.git/164794

Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agodemonstrate broken 'git cherry-pick three one two'
Martin von Zweigbergk [Wed, 29 Aug 2012 06:15:55 +0000 (23:15 -0700)] 
demonstrate broken 'git cherry-pick three one two'

Cherry-picking commits out of order (w.r.t. commit time stamp) doesn't
currently work. Add a test case to demonstrate it.

Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agoteach log --no-walk=unsorted, which avoids sorting
Martin von Zweigbergk [Wed, 29 Aug 2012 06:15:54 +0000 (23:15 -0700)] 
teach log --no-walk=unsorted, which avoids sorting

When 'git log' is passed the --no-walk option, no revision walk takes
place, naturally. Perhaps somewhat surprisingly, however, the provided
revisions still get sorted by commit date. So e.g 'git log --no-walk
HEAD HEAD~1' and 'git log --no-walk HEAD~1 HEAD' give the same result
(unless the two revisions share the commit date, in which case they
will retain the order given on the command line). As the commit that
introduced --no-walk (8e64006 (Teach revision machinery about
--no-walk, 2007-07-24)) points out, the sorting is intentional, to
allow things like

 git log --abbrev-commit --pretty=oneline --decorate --all --no-walk

to show all refs in order by commit date.

But there are also other cases where the sorting is not wanted, such
as

 <command producing revisions in order> |
       git log --oneline --no-walk --stdin

To accomodate both cases, leave the decision of whether or not to sort
up to the caller, by allowing --no-walk={sorted,unsorted}, defaulting
to 'sorted' for backward-compatibility reasons.

Signed-off-by: Martin von Zweigbergk <martinvonz@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agolog: fix --quiet synonym for -s
Jeff King [Tue, 28 Aug 2012 21:29:34 +0000 (17:29 -0400)] 
log: fix --quiet synonym for -s

Originally the "--quiet" option was parsed by the
diff-option parser into the internal QUICK option. This had
the effect of silencing diff output from the log (which was
not intended, but happened to work and people started to
use it). But it also had other odd side effects at the diff
level (for example, it would suppress the second commit in
"git show A B").

To fix this, commit 1c40c36 converted log to parse-options
and handled the "quiet" option separately, not passing it
on to the diff code. However, it simply ignored the option,
which was a regression for people using it as a synonym for
"-s". Commit 01771a8 then fixed that by interpreting the
option to add DIFF_FORMAT_NO_OUTPUT to the list of output
formats.

However, that commit did not fix it in all cases. It sets
the flag after setup_revisions is called. Naively, this
makes sense because you would expect the setup_revisions
parser to overwrite our output format flag if "-p" or
another output format flag is seen.

However, that is not how the NO_OUTPUT flag works. We
actually store it in the bit-field as just another format.
At the end of setup_revisions, we call diff_setup_done,
which post-processes the bitfield and clears any other
formats if we have set NO_OUTPUT. By setting the flag after
setup_revisions is done, diff_setup_done does not have a
chance to make this tweak, and we end up with other format
options still set.

As a result, the flag would have no effect in "git log -p
--quiet" or "git show --quiet".  Fix it by setting the
format flag before the call to setup_revisions.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agohttp: prompt for credentials on failed POST
Jeff King [Mon, 27 Aug 2012 13:27:15 +0000 (09:27 -0400)] 
http: prompt for credentials on failed POST

All of the smart-http GET requests go through the http_get_*
functions, which will prompt for credentials and retry if we
see an HTTP 401.

POST requests, however, do not go through any central point.
Moreover, it is difficult to retry in the general case; we
cannot assume the request body fits in memory or is even
seekable, and we don't know how much of it was consumed
during the attempt.

Most of the time, this is not a big deal; for both fetching
and pushing, we make a GET request before doing any POSTs,
so typically we figure out the credentials during the first
request, then reuse them during the POST. However, some
servers may allow a client to get the list of refs from
receive-pack without authentication, and then require
authentication when the client actually tries to POST the
pack.

This is not ideal, as the client may do a non-trivial amount
of work to generate the pack (e.g., delta-compressing
objects). However, for a long time it has been the
recommended example configuration in git-http-backend(1) for
setting up a repository with anonymous fetch and
authenticated push. This setup has always been broken
without putting a username into the URL. Prior to commit
986bbc0, it did work with a username in the URL, because git
would prompt for credentials before making any requests at
all. However, post-986bbc0, it is totally broken. Since it
has been advertised in the manpage for some time, we should
make sure it works.

Unfortunately, it is not as easy as simply calling post_rpc
again when it fails, due to the input issue mentioned above.
However, we can still make this specific case work by
retrying in two specific instances:

  1. If the request is large (bigger than LARGE_PACKET_MAX),
     we will first send a probe request with a single flush
     packet. Since this request is static, we can freely
     retry it.

  2. If the request is small and we are not using gzip, then
     we have the whole thing in-core, and we can freely
     retry.

That means we will not retry in some instances, including:

  1. If we are using gzip. However, we only do so when
     calling git-upload-pack, so it does not apply to
     pushes.

  2. If we have a large request, the probe succeeds, but
     then the real POST wants authentication. This is an
     extremely unlikely configuration and not worth worrying
     about.

While it might be nice to cover those instances, doing so
would be significantly more complex for very little
real-world gain. In the long run, we will be much better off
when curl learns to internally handle authentication as a
callback, and we can cleanly handle all cases that way.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agohttp: factor out http error code handling
Jeff King [Mon, 27 Aug 2012 13:26:04 +0000 (09:26 -0400)] 
http: factor out http error code handling

Most of our http requests go through the http_request()
interface, which does some nice post-processing on the
results. In particular, it handles prompting for missing
credentials as well as approving and rejecting valid or
invalid credentials. Unfortunately, it only handles GET
requests. Making it handle POSTs would be quite complex, so
let's pull result handling code into its own function so
that it can be reused from the POST code paths.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot: test http access to "half-auth" repositories
Jeff King [Mon, 27 Aug 2012 13:25:53 +0000 (09:25 -0400)] 
t: test http access to "half-auth" repositories

Some sites set up http access to repositories such that
fetching is anonymous and unauthenticated, but pushing is
authenticated. While there are multiple ways to do this, the
technique advertised in the git-http-backend manpage is to
block access to locations matching "/git-receive-pack$".

Let's emulate that advice in our test setup, which makes it
clear that this advice does not actually work.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot: test basic smart-http authentication
Jeff King [Mon, 27 Aug 2012 13:25:36 +0000 (09:25 -0400)] 
t: test basic smart-http authentication

We do not currently test authentication over smart-http at
all. In theory, it should work exactly as it does for dumb
http (which we do test). It does indeed work for these
simple tests, but this patch lays the groundwork for more
complex tests in future patches.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot/lib-httpd: recognize */smart/* repos as smart-http
Jeff King [Mon, 27 Aug 2012 13:25:21 +0000 (09:25 -0400)] 
t/lib-httpd: recognize */smart/* repos as smart-http

We do not currently test authentication for smart-http repos
at all. Part of the infrastructure to do this is recognizing
that auth/smart is indeed a smart-http repo.

The current apache config recognizes only "^/smart/*" as
smart-http. Let's instead treat anything with /smart/ in the
URL as smart-http. This is obviously a stupid thing to do
for a real production site, but for our test suite we know
that our repositories will not have this magic string in the
name.

Note that we will route /foo/smart/bar.git directly to
git-http-backend/bar.git; in other words, everything before
the "/smart/" is irrelevant to finding the repo on disk (but
may impact apache config, for example by triggering auth
checks).

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
12 years agot/lib-httpd: only route auth/dumb to dumb repos
Jeff King [Mon, 27 Aug 2012 13:24:42 +0000 (09:24 -0400)] 
t/lib-httpd: only route auth/dumb to dumb repos

Our test apache config points all of auth/ directly to the
on-disk repositories via an Alias directive. This works fine
because everything authenticated is currently in auth/dumb,
which is a subset.  However, this would conflict with a
ScriptAlias for auth/smart (which will come in future
patches), so let's narrow the Alias.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>