4 The latest copy of this document is found at
6 http://kernel.org/git/?p=git/git.git;a=blob;hb=todo;f=TODO
12 - In 0.99.8, we still install the backward compatible symbolic
13 links in $(bindir). These will however be removed before 1.0
16 git-ssh-push and git-ssh-pull pair is not going away within
17 this timeframe, if ever. Each of these old-name commands
18 continues to invoke its old-name counterpart on the other
22 What to expect after 0.99.8
23 ===========================
25 This is written in a form of to-do list for me, so if I say
26 "accept patch", it means I do not currently plan to do that
27 myself. People interested in seeing it materialize please take
28 a hint. Also whatever I marked "Perhaps" do not have to happen
29 if ever -- only if somebody cares enough and submits a clean
36 * Document the ref naming restrictions.
38 * David Ho's report suggests whatchanged documentation should
39 mention -m as "commonly used options". Steal Linus'
42 * Help Jon Loeliger to find place in the documentation to place
45 * Accept patches from people who actually have done CVS
46 migration and update the cvs-migration documentation.
47 Link the documentation from the main git.txt page.
49 * Talk about using rsync just once at the beginning when
50 initializing a remote repository so that local packs do not
51 need to be expanded. I personally do not think we need tool
52 support for this (but see below about optimized cloning).
54 * Maybe update tutorial with a toy project that involves two or
57 * Update tutorial to cover setting up repository hooks to do
60 * Do we still have missing docs? If so accept patches to finish
63 * Accept patches to talk about "Whoops, it broke. What's
66 * Accept patches to make formatted tables in asciidoc to work
67 well in both html and man pages (see git-diff(1)).
73 * We might want to optimize cloning with GIT native transport
74 not to explode the pack, and store it in objects/pack instead.
75 We would need a tool to generate an idx file out of a pack
76 file for this. Also this itself may turn out to be a bad
77 idea, making the set of packs in repositories everybody has
78 different from each other. [DONE; git-index-pack by Sergey,
81 * Git daemon, when deployed at kernel.org, might turn out to be
82 quite a burden, since it needs to generate customized packs
83 every time a new request comes in. It may be worthwhile to
84 precompute some packs for popular sets of heads downloaders
85 have and serve that, even if that could give more than the
86 client asks for in some cases. We will know about this soon
89 * Libification. There are many places "run once" mentality is
90 ingrained in the management of basic data structures, which
91 need to be fixed. [Matthias Urlichs is already working on
92 this: <pan.2005.10.03.20.48.52.132570@smurf.noris.de>; Post
95 * Maybe a pack optimizer.
97 Given a set of objects and a set of refs (probably a handful
98 branch heads and point release tags), find a set of packs to
99 allow reasonably minimum download for all of these classes of
100 people: (1) somebody cloning the repository from scratch, (2)
101 somebody who tends to follow the master branch head reasonably
102 closely, (3) somebody who tends to follow only the point
105 * Maybe an Emacs VC backend.
107 * 'git split-projects'? This requires updated 'git-rev-list' to
108 skip irrelevant commits.
109 Message-ID: <Pine.LNX.4.63.0509221617300.23242@iabervon.org>
111 * Look at libified GNU diff CVS seems to use, or libxdiff.
112 [Daniel has his own diff tool almost ready to start
113 integrating and testing; Post 1.0]
115 * Accept patches to fetch multiple objects by HTTP in parallel.
118 * Plug-in file-level merges [Post 1.0].
120 * Per-repository configuration mechanism [DONE by Linus].
126 * Merlyn reports trouble with http fetch.
128 * Maybe look at Cogito and see if I can help Pasky to adjust to
129 the later core features? Zack Brown's "cg-seek leaving empty
130 directories" problem is a good example of this.
132 * Perhaps detect cloning request in upload-pack and cache the
133 result for next cloning request until any of our refs change.
135 * Perhaps accept more "want"s in upload-pack and do something
136 intelligent about it.
138 * Perhaps send less "want"s from fetch-pack.
140 * Look at svn importer Smurf has. There is a small cvsimport
141 update in his tree as well [DONE].
143 * Decide the notation of "peeling the onion" operator, and
144 implement it in sha1_name.c. Perhaps postfix "^{}" to mean
145 "peel and expect anything", "^{blob}" to mean "peel and barf
146 unless blob". The current "^0" becomes shorthand for
149 * Quote the URL so that libcurl's metecharacter mechanism would
152 * Review the Makefile variables and exporting rules for them,
153 while looking at prefix passing by Kai Ruemmler [DONE].
155 * Review the 'sparse object database' change by Linus and move
156 the first phase of it to the "master" branch [DONE].
158 * Decide on mmap(). I am inclined to just stick to mmap
159 replacement by Johannes Schindelin and do nothing else right
160 now, except perhaps drop the writing-back support [DONE].
162 * Revisit Santi's patch to move commit temorary files out of the
163 working tree toplevel [DONE].
165 * More generally, review the use of temporary files again.
166 Assuming writable $GIT_DIR is more acceptable, but the
167 working tree toplevel may not be in a rare usage pattern.
169 * Encourage concrete proposals to commit log message templates
170 we discussed some time ago.
172 * Accept patches to cause "read-tree -u" delete a directory when
173 it makes it empty [DONE].
175 * Perhaps accept patches to do undo/redo.
177 * Perhaps accept patch to optionally allow '--fuzz' in
180 * Allow 'git apply' to accept GNU diff 2.7 output that forgets
181 to say '\No newline' if both input ends with incomplete
184 * What to do with TABs and LFs in pathnames without breaking GNU
185 patch? [DONE -- go with GNU patch extension proposed by Paul].
187 * Adjust apply.c to proposed GNU patch extension that quotes \n
188 and \t in C style, inside "". [DONE]
190 * Adjust diff.c to the same. [DONE]
192 * Maybe grok PGP signed text/plain in applymbox as well.
194 * Perhaps a tool to revert a single file to pre-modification
195 state? People with BK background know this operation as
196 'clean'. 'git checkout [-f] ent [path...]' was suggested by
197 Matthias Urlichs which sounds a natural extention to what the
198 command currently does.
200 * Enhance "git repack" to not always use --all; this would be
201 handy if the repository contains wagging heads like "pu" in
204 * Internally split the project into non-doc and doc parts; add
205 an extra root for the doc part and merge from it; move the
206 internal doc source to a separate repository, like the +Meta
207 repository; experiment if this results in a reasonable
208 workflow, and document it in howto form if it does.
210 The point is to make it possible to fork that part off to
211 somebody else; then I do not have to maintain Documentation
212 directory myself anymore, just like I simply slurp the latest
213 gitk from Paul and not worry about it ;-).
215 * Make rebase restartable; instead of skipping what cannot be
216 automatically forward ported, leave the conflicts in the work
217 tree, have the user resolve it, and then restart from where it
218 left off [mechanism mostly done].
220 * Output full path in the "git-rev-list --objects" output, not
221 just the basename, and see the improved clustering results in
222 better packing [Tried, but did not work out well].
224 * Updated git-changes-script Jeff Garzik needs [Inquiry for
225 external spec sent out with a quick hack. Will know if that
226 is what he needs hopefully soon].
228 * An mechanism to ignore filesystem mode bits altogether [DONE].
234 * Peter Hagervall's sparse fix [DONE].
236 * Alex Riesen reported that hooks are in effect in tests.
239 * Disallow [\001-\040\177] byte values from ref names. Also we
240 need to disallow ':' (used in refspec), '^' and '~' (postfix
241 "peel the onion" operators), and '..' ("ref1..ref2" notation
242 becomes ambiguous otherwise) [DONE].
244 * Update fetch-pack and clone-pack to ignore funny refs from the
245 other end, while making sure peek-remote does not discard them.
248 * Update upload-pack to send tag^{}. This would hopefully help
249 Pasky's automated tag tracking, and also Martin's findtags.
252 * Adjust update-index to quoted --index-info [DONE].
254 * Prepare apply.c changes for maint branch (0.99.8e) [READY].
256 * Perhaps show ^{commit}, ^{tree} instead of ^{} from ls-remote?
258 * Re-adjust maint branch for the above if we did so.
260 * short SHA1 naming is not enforcing uniqueness. Should fix [DONE].
262 * 'git repack' can be DOSed. Should fix [DONE].
264 * Stop installing the old-name symlinks [POSTPONED, but before 1.0].
266 * 'git merge-projects'?
268 Subject: Re: Merges without bases
269 References: <1125004228.4110.20.camel@localhost.localdomain>
270 Date: Thu, 25 Aug 2005 15:26:36 -0700
271 Message-ID: <7vvf1tps9v.fsf@assigned-by-dhcp.cox.net>
273 * 'git lost-and-found'? Link dangling commits found by
274 fsck-objects under $GIT_DIR/refs/lost-found/. Then
275 show-branch or gitk can be used to find any lost commit. [A
276 feeler patch sent out. Very underwhelming response X-<.]
278 Do not name it /lost+found/; that would probably confuse
279 things that mistake it a mount point (not our code but
282 * Add simple globbing rules to git-show-branch so that I can
283 say 'git show-branch --heads "ko-*"' (ko-master, ko-pu, and
284 ko-rc are in refs/tags/).
286 * We would want test scripts for the relative directory path
287 stuff Linus has been working on. So far, the following
288 commands should be usable with relative directory paths:
298 * In a freashly created empty repository, `git fetch foo:bar`
299 works OK, but `git checkout bar` afterwards does not (missing