While rpmbuild and mock have --nocheck to disable the %check section,
'fedpkg mockbuild' lacks this convenient option.
Add %bcond_without tests to allow 'fedpkg mockbuild --without tests' to
not run the test suite. Disabling the test suite cuts the build time by
approximately 60%, which is very useful while working on changes to the
packaging.
With curl-7.61.1 cookies are sorted by creation-time¹. Sort the output
used in the 'cookies stored in http.cookiefile when http.savecookies
set' test before comparing it to the expected cookies.
¹ https://github.com/curl/curl/commit/e2ef8d6fa ("cookies: support
creation-time attribute for cookies", 2018-08-28)
When building with options "--without docs --without p4 --without cvs"
the build fails with the following errors:
error: Installed (but unpackaged) file(s) found:
/usr/share/doc/git/git-cvsexportcommit.txt
/usr/share/doc/git/git-cvsimport.txt
/usr/share/doc/git/git-cvsserver.txt
/usr/share/doc/git/git-p4.txt
Installed (but unpackaged) file(s) found:
/usr/share/doc/git/git-cvsexportcommit.txt
/usr/share/doc/git/git-cvsimport.txt
/usr/share/doc/git/git-cvsserver.txt
/usr/share/doc/git/git-p4.txt
The .txt files were not caught by the %files entry in the main git
package when cvs/p4 were disabled -- from 9cd8ee7 ("Disable CVS support
on EL > 7", 2018-03-14. Those only applied when documentation was not
disabled.
Remove git-cvs* and git-p4* files from Documentation as well as the
%{buildroot}. Simplify the find path by dropping %{_bindir} and
%{gitexecdir}. Tighten the git-p4 glob to avoid unintended matches.
Drop the conditional inclusion of cvs/p4 docs in the main git package in
favor of removing the files entirely.
Gives possibility to add dependencies for git-instaweb http daemon,
without having to install all dependencies at each git install.
Currently, lighttpd is required by the git-instaweb package.
The git-instaweb script supports other httpd daemons (httpd, mongoose,
plackup [in perl(Plack)], and webrick [in rub-libs]). lighttpd is the
default, works without any configuration, and is only ~1M installed.
Add a conditional to allow merging from master to f29. The obsoletes
should be removed when f29 is EOL. It was added in 2d1c8b1 ("Remove
obsolete gnome-keyring credential helper", 2018-01-09). The comment was
improved in 4a06e99 ("clarify comment for obsolete git-gnome-keyring",
2018-09-04).
Avoid shipping scripts which require python2 when building without
python2. The following scripts/directories are removed:
contrib/fast-import/import-zips.py
contrib/hg-to-git
contrib/svn-fe
A future release of git will likely remove contrib/svn-fe and
git-remote-testsvn¹. The git-remote-testsvn binary is the only noarch
file in the git-svn package. Seeing that it's utility is very
questionable, remove it so git-svn can return to a noarch package.
¹ https://public-inbox.org/git/20180817190310.GA5360@sigill.intra.peff.net/
9125e65 ("Use new INSTALL_SYMLINKS setting", 2018-05-30) broke builds
using --without cvs. /usr/libexec/git-core/git-cvsserver became a
symlink instead of hardlink. Adapt the find command used to exclude
'git-cvs*' files to detect symlinks as well.
We want to build all documentation in the %build phase rather than
falling through to the %install phase and building it as a dependency of
install-doc.
The git-contacts script was added to SubmittingPatches recently. Make
it easier for users who read about it in the documentation to make use
of the command.
The default target in contrib/credential/netrc/Makefile is, and has
always been, test. Running 'make -C contrib/credential/netrc/' in
%build is not needed.
Additionally, the tests recently were changed and require perl-Git to be
installed before running. The tests also exit cleanly regardless of any
failures encountered, which makes them unreliable. A fix for these
issues will be submitted upstream, but rather than apply it here, simply
drop the unneeded 'make' call.
Ideally, the tests will be run in %check once fixed. This does present
a small wrinkle due to the deletion of contrib/credential in %install.
Cross that bridge when we get there. :)
Replace NO_CROSS_DIRECTORY_HARDLINKS and NO_INSTALL_HARDLINKS with
INSTALL_SYMLINKS. The result is slightly improved; all symlinks will
point directly to the target rather than via multiple levels of
symlinks.
The rationale was covered in slightly more detail in d56cfc6 ("Use
symlinks instead of hardlinks for installed binaries", 2018-03-15).
Adjust the dangling-relative-symlink filter in the rpmlint config for
the new target of the git-difftool symlink.
The USE_LIBPCRE setting now defaults to pcre2; use it. It's still
valid to set USE_LIBPCRE2, but using the default should be cleaner in
the long-run.
The (long-unmaintained) emacs support has been dropped upstream in favor
of better alternatives. From the upstream commit¹:
The git-blame.el mode has been superseded by Emacs's own
vc-annotate (invoked by C-x v g). Users of the git.el mode are now
much better off using either Magit or the Git backend for Emacs's own
VC mode.
These modes were added over 10 years ago when Emacs's own Git support
was much less mature, and there weren't other mature modes in the wild
or shipped with Emacs itself.
These days these modes have few if any users, and users of git aren't
well served by us shipping these (some OS's install them alongside git
by default, which is confusing and leads users astray).
¹ 6d5ed4836d ("git{,-blame}.el: remove old bitrotting Emacs code", 2018-04-11)
https://git.kernel.org/pub/scm/git/git.git/commit/?id=6d5ed4836d
Also drop DESTDIR and INSTALL from config.mak; they are both handled via
%make_install.
Remove the rpmlint filter for %buildroot usage which was only needed due
to DESTDIR's use in config.mak.
Specifically, t5512-ls-remote.sh has a test which starts a jgit daemon.
This has failed to exit on a number of occasions, only on s390x. We
could disable just that test with "GIT_SKIP_TESTS=t5512.28", but the
test number can and does change as more ls-remote tests are added.
Dropping the jgit BuildRequires is cleaner and only causes 3 tests to be
skipped, the offending t5512 test and two others in t5310-pack-bitmaps.
Access to s390x might help better debug this, but it does not occur
consistently and may be limited to koji. The issue could be a problem
in jgit as well. While looking at a hung build, Kevin Fenzi found a few
errors in t5512-ls-remote.out:
/usr/bin/build-classpath: Could not find xz-java Java extension for this JVM
/usr/bin/build-classpath: error: Some specified jars were not found
Unfortunately, it appears we need to carry this patch longer than
expected. Return to using %autosetup so other patches are easier to
manage. Use %apply_patch to manually apply the zlib patch only on
aarch64, as that is the only arch where it is required at this time.
A recent zlib build with optimization for ARM exposed an issue in git's
packfile handling.
Thanks to Pavel Cahyna for the initial report and debugging and Jeremy
Linton for further diagnosis and the subsequent patch.
The patch is currently being discussed upstream¹. Until it is accepted,
apply it only on aarch64 to avoid any unexpected issues with other
arches.
¹ https://public-inbox.org/git/20180525231713.23047-1-lintonrjeremy@gmail.com/T/#u
Fixes two security issues, described in the 2.13.7 release notes¹:
* Submodule "names" come from the untrusted .gitmodules file, but we
blindly append them to $GIT_DIR/modules to create our on-disk repo
paths. This means you can do bad things by putting "../" into the
name. We now enforce some rules for submodule names which will cause
Git to ignore these malicious names (CVE-2018-11235).
Credit for finding this vulnerability and the proof of concept from
which the test script was adapted goes to Etienne Stalmans.
* It was possible to trick the code that sanity-checks paths on NTFS
into reading random piece of memory (CVE-2018-11233).
¹ https://mirrors.edge.kernel.org/pub/software/scm/git/docs/RelNotes/2.13.7.txt
If 'make test' fails before running any tests, the debug output from
print-failed-test-output is confusing:
+ ./print-failed-test-output
cat: t/test-results/*.exit: No such file or directory
./print-failed-test-output: line 6: [: : integer expression expected
--------------------------------------------------------------------------------
t/test-results/*.out
--------------------------------------------------------------------------------
cat: t/test-results/*.out: No such file or directory
Use the bash failglob option to imrpve the output:
+ ./print-failed-test-output
./print-failed-test-output: line 12: no match: t/test-results/*.exit
The unknown, but temporary, breakage in fedora-28-x86_64 buildroots
appears to be resolved.
The test was disabled in a998227 ("Disable t5000-tar-tree.sh on x86 in
f28", 2018-01-18).
The spec file is a bit easier to read with as few conditional blocks as
required. Use %bcond_(with|without) to allow easier toggling of the
link checking.
