"git blame" learnt to take "--first-parent" and "--reverse" at the
same time when it makes sense.
* mk/blame-first-parent:
blame: allow blame --reverse --first-parent when it makes sense
blame: extract find_single_final
blame: test to describe use of blame --reverse --first-parent
"git status --branch --short" accessed beyond the constant string
"HEAD", which has been corrected.
* rs/wt-status-detached-branch-fix:
wt-status: use skip_prefix() to get rid of magic string length constants
wt-status: don't skip a magical number of characters blindly
wt-status: avoid building bogus branch name with detached HEAD
wt-status: exit early using goto in wt_shortstatus_print_tracking()
t7060: add test for status --branch on a detached HEAD
We peek objects from submodule's object store by linking it to the
list of alternate object databases, but the code to do so forgot to
correctly initialize the list.
* jk/initialization-fix-to-add-submodule-odb:
add_submodule_odb: initialize alt_odb list earlier
The code to prepare the working tree side of temporary directory
for the "dir-diff" feature forgot that symbolic links need not be
copied (or symlinked) to the temporary area, as the code already
special cases and overwrites them. Besides, it was wrong to try
computing the object name of the target of symbolic link, which may
not even exist or may be a directory.
* da/difftool:
difftool: ignore symbolic links in use_wt_file
Using the timestamp based criteria in "git branch --sort" did not
tiebreak branches that point at commits with the same timestamp (or
the same commit), making the resulting output unstable.
* kn/for-each-branch:
ref-filter: fallback on alphabetical comparison
mailinfo: fix passing wrong address to git_mailinfo_config
git_mailinfo_config() expects "struct mailinfo *". But in
setup_mailinfo(), "mi" is already "struct mailinfo *". &mi would make
it "struct mailinfo **" and git_mailinfo_config() would damage some
other memory when it assigns some value to mi->use_scissors.
This is caught by t4150.20. git_mailinfo_config() breaks
mi->name.alloc and makes strbuf_release() in clear_mailinfo() attempt
to free strbuf_slopbuf.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
wt-status: don't skip a magical number of characters blindly
Use the variable branch_name, which already has "refs/heads/" removed,
instead of blindly advancing in the ->branch string by 11 bytes. This
is safer and less magical.
Signed-off-by: Rene Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
wt-status: avoid building bogus branch name with detached HEAD
If we're on a detached HEAD then wt_shortstatus_print_tracking() takes
the string "HEAD (no branch)", translates it, skips the first eleven
characters and passes the result to branch_get(), which returns a bogus
result and accesses memory out of bounds in order to produce it.
Somehow stat_tracking_info(), which is passed that result, does the
right thing anyway, i.e. it finds that there is no base.
Avoid the bogus results and memory accesses by checking for HEAD first
and exiting early in that case. This fixes t7060 with --valgrind.
Signed-off-by: Rene Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t7060: add test for status --branch on a detached HEAD
This test fails when run under Valgrind because branch_get() gets passed
a bogus branch name pointer:
==62831== Invalid read of size 1
==62831== at 0x4F76AE: branch_get (remote.c:1650)
==62831== by 0x53499E: wt_shortstatus_print_tracking (wt-status.c:1654)
==62831== by 0x53499E: wt_shortstatus_print (wt-status.c:1706)
==62831== by 0x428D29: cmd_status (commit.c:1384)
==62831== by 0x405D6D: run_builtin (git.c:350)
==62831== by 0x405D6D: handle_builtin (git.c:536)
==62831== by 0x404F10: run_argv (git.c:582)
==62831== by 0x404F10: main (git.c:690)
==62831== Address 0x5e89b0b is 6 bytes after a block of size 5 alloc'd
==62831== at 0x4C28C4F: malloc (vg_replace_malloc.c:299)
==62831== by 0x59579E9: strdup (strdup.c:42)
==62831== by 0x52E108: xstrdup (wrapper.c:43)
==62831== by 0x5322A6: wt_status_prepare (wt-status.c:130)
==62831== by 0x4276E0: status_init_config (commit.c:184)
==62831== by 0x428BB8: cmd_status (commit.c:1350)
==62831== by 0x405D6D: run_builtin (git.c:350)
==62831== by 0x405D6D: handle_builtin (git.c:536)
==62831== by 0x404F10: run_argv (git.c:582)
==62831== by 0x404F10: main (git.c:690)
Signed-off-by: Rene Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
blame: allow blame --reverse --first-parent when it makes sense
Allow combining --reverse and --first-parent if initial commit of
specified range is at the first-parent chain starting from the final
commit. Disable the prepare_revision_walk()'s builtin children
collection, instead picking only the ones which are along the first
parent chain.
Signed-off-by: Max Kirillov <max@max630.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
blame: test to describe use of blame --reverse --first-parent
Reverse blame can be used to locate removal of lines which does not
change adjacent lines. Such edits do not appear in non-reverse blame,
because the adjacent lines last changed commit is older history, before
the edit.
For a big and active project which uses topic branches, or analogous
feature, for example pull-requests, the history can contain many
concurrent branches, and even after an edit merged into the target
branch, there are still many (sometimes several tens or even hundreds)
topic branch which do not contain it:
Here, the '*'s mark the first parent in merge, and uppercase B1 - the
commit where the line being blamed for was removed. Since commits cN-zN
do not contain B1, they still have the line removed in B1, and
reverse blame can report that the last commit for the line was zN
(meaning that it was removed in a100). In fact it really does return
some very late commit, and this makes it unusable for finding the B1
commit.
The search could be done by blame --reverse --first-parent. For range
a0..a100 it would return a1, and then only one additional blame along
the a0..bN will return the desired commit b0. But combining --reverse
and --first-parent was forbidden in 95a4fb0eac, because incorrectly
specified range could produce unexpected and meaningless result.
Add test which describes the expected behavior of
`blame --reverse --first-parent` in the case described above.
Signed-off-by: Max Kirillov <max@max630.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
test: facilitate debugging Git executables in tests with gdb
When prefixing a Git call in the test suite with 'debug ', it will
now be run with GDB, allowing the developer to debug test failures
more conveniently.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In ref-filter.c the comparison of refs while sorting is handled by
cmp_ref_sorting() function. When sorting as per numerical values
(e.g. --sort=objectsize) there is no fallback comparison when both
refs hold the same value. This can cause unexpected results (i.e. the
order of listing refs with equal values cannot be pre-determined) as
pointed out by Johannes Sixt ($gmane/280117).
Hence, fallback to alphabetical comparison based on the refname
whenever the other criterion is equal.
A test in t3203 was expecting that branch-two sorts before HEAD, which
happened to be how qsort(3) on Linux sorted the array, but (1) that
outcome was not even guaranteed, and (2) once we start breaking ties
with the refname, "HEAD" should sort before "branch-two" so the
original expectation was inconsistent with the criterion we now use.
Update it to match the new world order, which we can now depend on
being stable.
Helped-by: Junio C Hamano <gitster@pobox.com> Reported-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Karthik Nayak <Karthik.188@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Recent update to "rebase -i" that tries to sanity check the edited
insn sheet before it uses it has become too picky on Windows where
CRLF left by the editor is turned into a trailing CR on the line
read via the "read" built-in command.
* gr/rebase-i-drop-warn:
rebase-i: work around Windows CRLF line endings
t3404: "rebase -i" gets broken when insn sheet uses CR/LF line endings
Various compilation fixes and squelching of warnings.
* js/misc-fixes:
Correct fscanf formatting string for I64u values
Silence GCC's "cast of pointer to integer of a different size" warning
Squelch warning about an integer overflow
Merging a branch that removes a path and another that changes the
mode bits on the same path should have conflicted at the path, but
it didn't and silently favoured the removal.
* jk/delete-modechange-conflict:
merge: detect delete/modechange conflict
t6031: generalize for recursive and resolve strategies
t6031: move triple-rename test to t3030
"git clone --dissociate" used to require that "--reference" was
used at the same time, but you can create a new repository that
borrows objects from another without using "--reference", namely
with "clone --local" from a repository that borrows objects from
other repositories.
* ar/clone-dissociate:
clone: allow "--dissociate" without reference
gitk: Update msgid's for menu items with accelerator
The commit d99b4b0de27a ("gitk: Accelerators for the main menu")
modified the menu item strings with the accelerator, but the
translations didn't follow, thus the menus are shown without
translations.
This patch systematically update the msgid keys just to follow this
change. The contents aren't changed, so the accelerator won't work in
these locales for now. Each locale translator needs to add proper
acceleration keys appropriately.
Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Paul Mackerras <paulus@samba.org>
In d99b4b0de27a ("gitk: Accelerators for the main menu", 2015-09-09),
accelerators were added to allow efficient keyboard navigation. One
instance of the strings "Edit view..." and "Delete view" were left
without the ampersand.
Add the missing ampersand characters to unbreak our international
users.
Signed-off-by: Beat Bolli <dev+git@drbeat.li> Signed-off-by: Paul Mackerras <paulus@samba.org>
The implementation of "git mailinfo" was refactored so that a
mailinfo() function can be directly called from inside a process.
* jc/mailinfo-lib: (34 commits)
mailinfo: remove calls to exit() and die() deep in the callchain
mailinfo: handle charset conversion errors in the caller
mailinfo: libify
mailinfo: keep the parsed log message in a strbuf
mailinfo: handle_commit_msg() shouldn't be called after finding patchbreak
mailinfo: move content/content_top to struct mailinfo
mailinfo: move [ps]_hdr_data to struct mailinfo
mailinfo: move cmitmsg and patchfile to struct mailinfo
mailinfo: move charset to struct mailinfo
mailinfo: move transfer_encoding to struct mailinfo
mailinfo: move check for metainfo_charset to convert_to_utf8()
mailinfo: move metainfo_charset to struct mailinfo
mailinfo: move use_scissors and use_inbody_headers to struct mailinfo
mailinfo: move add_message_id and message_id to struct mailinfo
mailinfo: move patch_lines to struct mailinfo
mailinfo: move filter/header stage to struct mailinfo
mailinfo: move global "FILE *fin, *fout" to struct mailinfo
mailinfo: move keep_subject & keep_non_patch_bracket to struct mailinfo
mailinfo: introduce "struct mailinfo" to hold globals
mailinfo: move global "line" into mailinfo() function
...
The name-hash subsystem that is used to cope with case insensitive
filesystems keeps track of directories and their on-filesystem
cases for all the paths in the index by holding a pointer to a
randomly chosen cache entry that is inside the directory (for its
ce->ce_name component). This pointer was not updated even when the
cache entry was removed from the index, leading to use after free.
This was fixed by recording the path for each directory instead of
borrowing cache entries and restructuring the API somewhat.
* dt/name-hash-dir-entry-fix:
name-hash: don't reuse cache_entry in dir_entry
* xf/user-manual-markup:
Documentation: match undefline with the text in old release notes
Documentation: match underline with the text
Documentation: fix header markup
The caller is preparing a narrowed-down copy of the working tree and
this function is asked if the path should be included in that copy.
If we say yes, the path from the working tree will be either symlinked
or copied into the narrowed-down copy.
For any path that is a symbolic link, the caller later fixes up the
narrowed-down copy by unlinking the path and replacing it with a
regular file it writes out that mimics the way how "git diff"
compares symbolic links.
Let's answer "no, you do not want to copy/symlink the working tree
file" for all symbolic links from this function, as we know the
result will not be used because it will be overwritten anyway.
Incidentally, this also stops the function from feeding a symbolic
link in the working tree to hash-object, which is a wrong thing to
do to begin with. The link may be pointing at a directory, or worse
may be dangling (both would be noticed as an error). Even if the
link points at a regular file, hashing the contents of a file that
is pointed at by the link is not correct (Git hashes the contents of
the link itself, not the pointee).
Signed-off-by: David Aguilar <davvid@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Git-merge-file is documented to return one of three exit
codes:
- zero means the merge was successful
- a negative number means an error occurred
- a positive number indicates the number of conflicts
Unfortunately, this all gets stuffed into an 8-bit return
code. Which means that if you have 256 conflicts, this wraps
to zero, and the merge appears to succeed (and commits a
blob full of conflict-marker cruft!).
This patch clamps the return value to a maximum of 127,
which we should be able to safely represent everywhere. This
also leaves 128-255 for other values. Shells (and some parts
of git) will typically represent signal death as 128 plus
the signal number. And negative values are typically coerced
to an 8-bit unsigned value (so "return -1" ends up as 255).
Technically negative returns have the same problem (e.g.,
"-256" wraps back to 0), but this is not a problem in
practice, as the only negative value we use is "-1".
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Editors on Windows can and do save text files with CRLF line
endings, which is the convention on the platform. We are seeing
reports that the "read" command in a port of bash to the environment
however does not strip the CRLF at the end, not adjusting for the
same convention on the platform.
This breaks the recently added sanity checks for the insn sheet fed
to "rebase -i"; instead of an empty line (hence nothing in $command),
the script was getting a lone CR in there.
Special case a lone CR and treat it the same way as an empty line to
work this around.
This patch (also) passes the test with Git for Windows, where the
issue was seen first.
Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
add_submodule_odb: initialize alt_odb list earlier
The add_submodule_odb function tries to add a submodule's
object store as an "alternate". It needs the existing list
to be initialized (from the objects/info/alternates file)
for two reasons:
1. We look for duplicates with the existing alternate
stores, but obviously this doesn't work if we haven't
loaded any yet.
2. We link our new entry into the list by prepending it to
alt_odb_list. But we do _not_ modify alt_odb_tail.
This variable starts as NULL, and is a signal to the
alt_odb code that the list has not yet been
initialized.
We then call read_info_alternates on the submodule (to
recursively load its alternates), which will try to
append to that tail, assuming it has been initialized.
This causes us to segfault if it is NULL.
This rarely comes up in practice, because we will have
initialized the alt_odb any time we do an object lookup. So
you can trigger this only when:
- you try to access a submodule (e.g., a diff with
diff.submodule=log)
- the access happens before any other object has been
accessed (e.g., because the diff is between the working
tree and the index)
- the submodule contains an alternates file (so we try to
add an entry to the NULL alt_odb_tail)
To fix this, we just need to call prepare_alt_odb at the
start of the function (and if we have already initialized,
it is a noop).
Note that we can remove the prepare_alt_odb call from the
end. It is guaranteed to be a noop, since we will have
called it earlier.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prepare for Git on-disk repository representation to undergo
backward incompatible changes by introducing a new repository
format version "1", with an extension mechanism.
* jk/repository-extension:
introduce "preciousObjects" repository extension
introduce "extensions" form of core.repositoryformatversion
The internal stripspace() function has been moved to where it
logically belongs to, i.e. strbuf API, and the command line parser
of "git stripspace" has been updated to use the parse_options API.
* tk/stripspace:
stripspace: use parse-options for command-line parsing
strbuf: make stripspace() part of strbuf
A couple of commands still showed "[options]" in their usage string
to note where options should come on their command line, but we
spell that "[<options>]" in most places these days.
* rt/placeholder-in-usage:
am, credential-cache: add angle brackets to usage string
The synopsis text and the usage string of subcommands that read
list of things from the standard input are often shown as if they
only take input from a file on a filesystem, which was misleading.
* jc/usage-stdin:
usage: do not insist that standard input must come from a file
* mr/worktree-list:
worktree: add 'list' command
worktree: add details to the worktree struct
worktree: add a function to get worktree details
worktree: refactor find_linked_symref function
worktree: add top-level worktree.c
"git am -3" had a small regression where it is aborted in its error
handling codepath when underlying merge-recursive failed in certain
ways, as it assumed that the internal call to merge-recursive will
never die, which is not the case (yet).
* jc/am-3-fallback-regression-fix:
am -3: do not let failed merge from completing the error codepath
If one side deletes a file and the other changes its
content, we notice and report a conflict. However, if
instead of changing the content, we change only the mode,
the merge does not notice (and the mode change is silently
dropped).
The trivial index merge notices the problem and correctly
leaves the conflict in the index, but both merge-recursive
and merge-one-file will silently resolve this in favor of
the deletion. In many cases that is a sane resolution, but
we should be punting to the user whenever there is any
question. So let's detect and treat this as a conflict (in
both strategies).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The t6031 test was introduced to check filemode handling of
merge-recursive. Much later, an unrelated test was tacked on
to look at renames and d/f conflicts. This test does not
depend on anything that happened before (it actually blows
away any existing content in the test repo). Let's move it
to t3030, where there are more related tests.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This fix is probably purely cosmetic because PRIuMAX is likely identical
to SCNuMAX. Nevertheless, when using a function of the scanf() family,
the correct interpolation to use is the latter, not the former.
Signed-off-by: Waldek Maleska <w.maleska@gmail.com> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Silence GCC's "cast of pointer to integer of a different size" warning
When calculating hashes from pointers, it actually makes sense to cut
off the most significant bits. In that case, said warning does not make
a whole lot of sense.
So let's just work around it by casting the pointer first to intptr_t
and then casting up/down to the final integral type.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "Fast-forward merges" section of user-manual.txt incorrectly
says if the current branch is a descendant of the other, Git will
perform a fast-forward merge, but it should the other way around.
Signed-off-by: Xue Fuqiao <xfq.free@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since Git 2.0, "add -u" and "add -A" run from a subdirectory without
any pathspec mean "everything in the working tree" (before 2.0, they
were limited to the current directory). The limiting to the current
directory was implemented by inserting "." to the command line when
the end user did not give us any pathspec. At 2.0, we updated the
code to insert ":/" (instead of '.') to consider everything from the
top-level, by using a pathspec magic "top".
The call to parse_pathspec() using the command line arguments is,
however, made with PATHSPEC_PREFER_FULL option since 5a76aff1 (add:
convert to use parse_pathspec, 2013-07-14), which predates Git 2.0.
In retrospect, there was no need to turn "adding . to limit to the
directory" into "adding :/ to unlimit to everywhere" in Git 2.0;
instead we could just have done "if there is no pathspec on the
command line, just let it be". The parse_pathspec() then would give
us a pathspec that matches everything and all is well.
Incidentally such a simplification also fixes a corner case bug that
stems from the fact that ":/" does not necessarily mean any magic.
A user would say "git --literal-pathspecs add -u :/" from the
command line when she has a directory ':' and wants to add
everything in it (and she knows that her :/ will be taken as
'everything under the sun' magic pathspec unless she disables the
magic with --literal-pathspecs). The internal use of ':/' would
behave the same way as such an explicitly given ":/" when run with
"--literal-pathspecs", and will not add everything under the sun as
the code originally intended.
5096d490 (convert trivial sprintf / strcpy calls to xsnprintf) converted
two sprintf calls. Now GCC warns that "format '%u' expects argument of
type 'unsigned int', but argument 4 has type 'long unsigned int'".
Instead of changing the format string, use a variable of type unsigned
in place of the typedef-ed type DWORD, which hides that it is actually an
unsigned long.
There is no correctness issue with the old code because unsigned long and
unsigned are always of the same size on Windows, even in 64-bit builds.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The earlier rewrite f28e3ab2 (read_branches_file: simplify string handling)
of read_branches_file() lost an fclose() call. Put it back.
As on Windows files that are open cannot be removed, the leak manifests in
a failure of 'git remote rename origin origin' when the remote's URL is
specified in .git/branches/origin, because by the time that the command
attempts to remove this file, it is still open.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: AsciiDoc spells em-dash as double-dashes, not triple
Again, we do not usually process release notes with AsciiDoc, but it
is better to be consistent.
This incidentally reveals breakages left by an ancient 5e00439f
(Documentation: build html for all files in technical and howto,
2012-10-23). The index-format documentation was originally written
to be read as straight text without formatting and when the commit
forced everything in Documentation/ to go through AsciiDoc, it did
not do any adjustment--hence the double-dashes will be seen in the
resulting text that is rendered as preformatted fixed-width without
converted into em-dashes.
The "--reference" option is not the only way to provide a repository
to borrow objects from. A repository that borrows from another
repository can be cloned with "clone --local" and the resulting
repository will borrow from the same repository, which the user
may want to "--dissociate" from.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
After switching to use the tempfile module in commit 6e122b44
(setup_temporary_shallow(): use tempfile module), no declarations from
sigchain.h are used in read-cache.c anymore. Thus, remove the #include.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
After switching to use the tempfile module in commit f6ecc62d
(write_shared_index(): use tempfile module), no declarations from
sigchain.h are used in read-cache.c anymore. Thus, remove the #include.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
After switching to use the tempfile module in commit 284098f1
(diff: use tempfile module), no declarations from sigchain.h are used in
diff.c anymore. Thus, remove the #include.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
After switching to use the tempfile module in commit 9e903316
(credential-cache--daemon: use tempfile module), no declarations from
sigchain.h are used in credential-cache--daemon.c anymore. Thus, remove
the #include.
Signed-off-by: Tobias Klauser <tklauser@distanz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Even though AsciiDoc is more lenient when deciding if an underline
is for the contents on the previous line to find section headers, we
should match the length of them for other formatters to help them.
Documentation/everyday: match undefline with the text
Even though AsciiDoc is more lenient when deciding if an underline
is for the contents on the previous line to find section headers, we
should match the length of them for other formatters to help them.
Asciidoctor is stricter than AsciiDoc when deciding if underlining
is a section title or the start of preformatted text. Make the
length of the underlining match the text to ensure that it renders
correctly in all implementations.
Signed-off-by: Xue Fuqiao <xfq.free@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
And finally the endgame. Instead of spawning "git mailinfo" via the
run_command() API the same number of times as there are incoming
patches, make direct internal call to the libified mailinfo() from
"git am" to reduce the spawning overhead, which would matter on some
platforms.
mailinfo: handle charset conversion errors in the caller
Instead of dying in convert_to_utf8(), just report an error and let
the callers handle it. Between the two callers:
- decode_header() silently punts when it cannot parse a broken
RFC2047 encoded text (e.g. when it sees anything other than B or
Q after it sees "=?<charset>") by jumping to release_return,
returning the string it successfully parsed out so far, to the
caller. A piece of string that convert_to_utf8() cannot handle
can be treated the same way.
- handle_commit_msg() doesn't cope with a malformed line well, so
die there for now. We'll lift this even higher in later changes
in this series.
When mailinfo() is eventually libified, the calling "git am" still
will have to write out the log message in the "msg" file for hooks
and other users of the information, but it does not have to reopen
and reread what it wrote earlier if the function kept it in a strbuf.
This also removes the need for seeking and truncating the output
file when we see a scissors mark in the input, which in turn allows
us to lose two callsites of die_errno().
mailinfo: handle_commit_msg() shouldn't be called after finding patchbreak
There is a strange "if (!mi->cmitmsg) return 0" at the very beginning
of handle_commit_msg(), but the condition should never trigger, because:
* The only place cmitmsg is set to NULL is after this function sees
a patch break, closes the FILE * to write the commit log message
and returns 1. This function returns non-zero only from that
codepath.
* The caller of this function, upon seeing a non-zero return,
increments filter_stage, starts treating the input as patch text
and will never call handle_commit_msg() again.
Replace it with an assert(!mi->filter_stage) to ensure the above
observation will stay to be true.
mailinfo: move check for metainfo_charset to convert_to_utf8()
All callers of this function refrain from calling it when
mi->metainfo_charset is NULL; move the check to the callee,
as it already has a few conditions at its beginning to turn
it into a no-op.