Hi everyone,
Finally, the time has come, to sum up all my work so far into a single blog post. Without wasting any time, let’s get started.
WHEN I MET GIT?
I was planning to contribute in Git since Aug’19. And I made my first contribution to Git via Gitgitgadget in Oct’19. At that time I was mentored by dscho(aka Johannes). I continued to contribute even since then and managed to get some of my patches merged.
PRE-GSOC
Here’s the list of patches I worked on before GSoC.
-
get_main_worktree(): allow it to be called in the Git directory
-
git-compat-util.h: drop the
PRIuMAXand other fallback definitions
Note: These patches was considered as my micro-project.
PROJECT
Title: Unify ref-filter formats with other –pretty formats
Aim: To simplify the codebase by removing duplicate logic. The task is to re-use ref-filter.{c,h}’s logic in pretty.{c,h}.
Mentors: Christian Couder, Heba Waly
Note - For more details, you can check out my proposal here.
PROGRESS
Almost all of the pretty formats are implemented. I’m only mentioning the list of formatting options those are using ref-filter’s logic.
%H : commit hash
%h : abbreviated commit hash
%T : tree hash
%t : abbreviated tree hash
%P : parent hashes
%p : abbreviated parent hashes
%an : author name
%aN : author name (respecting .mailmap)
%ae : author email
%aE : author email (respecting .mailmap)
%al : author email local-part (the part before the @ sign)
%aL : author local-part (respecting .mailmap)
%ad : author date
%aD : author date, RFC2822 style
%ar : author date, relative
%at : author date, UNIX timestamp
%ai : author date, ISO 8601-like format
%aI : author date, strict ISO 8601 format
%as : author date, short format (YYYY-MM-DD)
%cn : committer name
%cN : committer name (respecting .mailmap)
%ce : committer email
%cE : committer email (respecting .mailmap)
%cl : author email local-part (the part before the @ sign)
%cL : committer local-part (respecting .mailmap)
%cd : committer date
%cD : committer date, RFC2822 style
%cr : committer date, relative
%ct : committer date, UNIX timestamp
%ci : committer date, ISO 8601-like format
%cI : committer date, strict ISO 8601 format
%cs : committer date, short format (YYYY-MM-DD)
%s : subject
%f : sanitized subject line, suitable for a filename
%b : body
%B : raw body (unwrapped subject and body)
%(trailers[:options]) : "only[=val]", "unfold[=val]", "valueonly[=val]"l, "key=<K>" and "separator=<SEP>"
Commit formats:
- oneline
- short
- default
- medium
- full
- fuller
- raw
WHATS LEFT?
Although we have implemented all the formatting options in the new file pretty-lib.{c,h}, it’s still not perfect. Some issues need work.
- Around 30% log related tests are failing.
I ran tests on the branch pretty-lib-2.0.0. This branch doesn’t consist of mailmap logic and mbox/email commit format. Some tests are failing for that reason. Also, it doesn’t handle unknown formatting options, some test might be failing because of this. And some other tests are failing for some unknown reason.
- Teach pretty-lib.{c,h} to handle incorrect formatting options.
Initially, we used die() in case of an unknown formatting option. But that’s not how original pretty.{c,h} handles unknown options. But recently I changed it to return 0 in case of unknown formatting option, just like pretty.c. It is giving segmentation fault in case of unknown formatting option. I didn’t get much time to investigate, but it seems like the caller function doesn’t know how to handle return 0 properly.
For ex: git log --pretty="%xy" gives Segmentation Fault.
Branch used for testing: pretty-lib-2.0.2
- EMAIL/MBOXED commit format needs work.
Logs can also be viewed as mbox format. This commit formatting option respects mailmap. We implemented mailmap in ref-filter a month ago. But as we modified new email options after Junio’s review (see here). We might also need to modify the mailmap logic. This also impacts mbox/email commit format.
Note: It works fine with old branches. But after we made changes to newly added email options logic, it is not possible to rebase mailmap patches without modifying them.
- Olga’s work also needs attention.
Olga’s work was not very close to mine but still is somewhat similar. I thought of working on Olga’s leftovers, but due to time constraints, I wasn’t able to do that. I might continue working on her patches sometime in the future. If you are reading this, you might be interested in Olga’s work too. Here are some links to her works.
Blog Series: https://medium.com/@olyatelezhnaya
Commits: https://github.com/telezhnaya/git/commits/format
I plan to keep on working on these issues even after GSoC ends. I tend to finish What i started.
PATCHES
Patch series sent to mailing list:
[PATCH 0/8] Improvements to ref-filter
* hv/ref-filter-misc (2020-08-17) 9 commits
(merged to 'next' on 2020-08-27 at c015fa6b0f)
+ ref-filter: add `sanitize` option for 'subject' atom
+ format-support: move `format_sanitized_subject()` from pretty
+ pretty: refactor `format_sanitized_subject()`
+ ref-filter: add `short` modifier to 'parent' atom
+ ref-filter: add `short` modifier to 'tree' atom
+ ref-filter: rename `objectname` related functions and fields
+ ref-filter: modify error messages in `grab_objectname()`
+ ref-filter: refactor `grab_objectname()`
+ ref-filter: support different email formats
The "--format=" option to the "for-each-ref" command and friends
learned a few more tricks, e.g. the ":short" suffix that applies to
"objectname" now also can be used for "parent", "tree", etc.
Will merge to 'master'.
Link to patch series and discussion on mailing list: https://public-inbox.org/git/pull.684.v4.git.1598046110.gitgitgadget@gmail.com/
Gitgitgadget PR: https://github.com/gitgitgadget/git/pull/684
[PATCH 0/2] Fix trailers atom bug and improved tests
* hv/ref-filter-trailers-atom-parsing-fix (2020-08-21) 2 commits
(merged to 'next' on 2020-08-24 at 79b27f3263)
+ ref-filter: 'contents:trailers' show error if `:` is missing
+ t6300: unify %(trailers) and %(contents:trailers) tests
The parser for "git for-each-ref --format=..." was too loose when
parsing the "%(trailers...)" atom, and forgot that "trailers" and
"trailers:<modifers>" are the only two allowed forms, which has
been corrected.
Will merge to 'master'.
Link to patch series and discussion on mailing list: https://public-inbox.org/git/pull.707.git.1597841551.gitgitgadget@gmail.com/
Gitgitgadget PR: https://github.com/gitgitgadget/git/pull/707
[PATCH 0/2] Unify trailers logic for pretty.{c,h} and ref-filter.{c,h}
- pretty.c: refactor trailer logic to `format_set_trailers_options()
- ref-filter: using pretty.c logic for trailers
Status: Waiting for another patch series to be merged to next.
[PATCH 0/30] pretty use ref-filter’s logic
- pretty-lib: add ‘raw’ commit format
- pretty-lib: print indented commits
- pretty: rename and make some generic support functions public
- ref_format: change
need_color_reset_at_eolto bit field - pretty-lib: add support for
--log-size - pretty-lib: add support for
--graph - pretty-lib: add formatting option
%<and%> - pretty: make
parse_padding_placeholder()public - pretty-lib: add formatting option
%w - pretty: move line wrapping logic to new function
- pretty-lib: add formatting option
%S - pretty-lib: add formatting option
%e - pretty-lib: add formatting option
%N - pretty-lib: support options that expands to string literals
- pretty-lib: add reflog related formatting options
- pretty: move reflog formatting logic to new function
- pretty-lib: add
raw bodyformatting option - pretty-lib: add different date formats
- pretty-lib: add formatting option
m,d, andD - pretty-lib: add color related formatting options
- pretty: make
struct format_commit_contextpublic - pretty.c: move color logic to
format_commit_color() - pretty-lib: add formatting option
f - pretty-lib: add formatting option
tandp - pretty-lib: add more formatting option
- pretty-lib: print commits using ref-filters logic
- ref_format: add option to skip
\nat eol - pretty: introduce
get_user_format() - revision: add
use_ref_filterin struct rev_info - builtin/log: new config log.useRefFilter
Link to commits: https://github.com/harry-hov/git/commits/pretty-lib-2.0.2?author=harry-hov
Status: Failing tests.
Some Important Branches
Here’s the list of some branches that contain above patches and some unfinished work.
- Branch without new file format-support.{c,h}:
pretty-lib-2.0.2
Why it exists: Because junio thinks that there is no point in adding new format-support.{c,h} if we are only making some pretty.{c,h}’s static functions public.
- Branch with new file format-support.{c,h}:
fix-graph3
Why it exists: Initially, we thought it would be nice to have another pair of files to keep the functions that we used in both pretty.c and ref-filter.c.
- Branch with new signature atom for ref-format:
cc-signature2
Why it exists: We implemented this new signature atom just like we initially introduced new email formats to ref-filter. We might need to refactor the logic Junio’s way before rebasing it with branch pretty-lib-2.0.2.
CLOSING REMARKS
Overall it was a great experience working with a Git and with my mentors. I intend to keep on contributing even after GSOC period ends. I would like to thank some of the awesome people I met during my journey with Git and GSOC.
- Christian Couder
- Heba Waly
- Eric Sunshine
- Johannes Schindelin
- Junio C. Hamano
Ciao.
Thanks for reading!
Hariom
