Age | Commit message (Collapse) | Author | Files | Lines |
|
Preallocating for very small fails causes some filesystems to degrade
performance. Do so only for "big enough" files.
Closes: #824636
|
|
On unsupported or exotic systems where we might be unable to retrieve
the number of online processors and have been requested to infer them
ourselves via the --jobs or --jobs-try “auto” argument, default to
serial execution instead of unlimited jobs, to be on the safe side.
Proposed-by: Simon McVittie <smcv@debian.org>
|
|
|
|
Using undeterministic filenames based on the buildinfo-id produces ugly
looking filenames, which get left behind when rebuilding the same source
multiple times as they vary by date.
There's really no great point in using unique filenames as they will end
up with different contents depending on the builder.
|
|
This records the time the build happened. This might be useful when
there is a need to track down problems caused by external time-based
events not visible from inside the build system. Things like hardware,
software deployment or other such failures.
|
|
|
|
|
|
This function only works with real and known Debian architectures. It
will not work with wildcards, nor the special architectures 'all' nor
'source'. Validate the arguments and croak early on to spot bogus usage.
Prompted-by: Johannes Schauer <josch@debian.org>
|
|
It seems like at least glibc does not support statically linked PIE
programs, and produces random junk. Disable globally for now, if
there's a desire to enable static PIE binaries, which is known to
work on some architectures (such as musl-based ones) we can add
specialized specs files in the future.
Closes: #843714
Proposed-by: Szabolcs Nagy <nsz@port70.net>
|
|
This makes sure we will not forget to get them out of the ignore list
once they get documented.
|
|
Update to 2823t1f.
|
|
|
|
|
|
|
|
Explicitly mention in dpkg-buildpackage(1) that setting the number
of parallel jobs to 1, restores the serial behavior.
Prompted-by: Thorsten Glaser <t.glaser@tarent.de>
|
|
Mention that this flag might also affect the beahavior of debian/rules,
and also mention the parallel option.
|
|
This will allow other projects to use the same whitelist as dpkg does.
Requested-by: Johannes Schauer <josch@debian.org>
|
|
When a broken date is on May the error message ended up being very
confusing as it mentioned that it was using a full instead of an
abbreviated month name, because for May both are the same.
Nest the conditionals to avoid this problem.
Closes: #843829
Signed-off-by: Guillem Jover <guillem@debian.org>
|
|
Our test run with the setting activated by apt, has not shown any signs
of pitch forks nor angry mobs. Let's switch this by default to auto, so
that the more clear but still hopefully unobtrusive colors can help
people deal with interesting messages, warnings or error conditions.
|
|
We are releasing the dpkg database now after running the commands, which
means that the postponed error reporting summary was trying to print
messages that had already been freed from the database memory pool.
Duplicate the passed strings so that we are impervious to the database
life-cycle.
Regression introduced in commit 3404fd24ef8020b4d6dc17adb82d7e6c035d90dc.
Closes: #843874
|
|
- Rename the spec name cc1_options to self_spec.
- Do not set PIE options if they have been negated, and do not reset
them if they have been requested.
Closes: #843791, #843826
|
|
As the former is more portable, and we were using GNUisms.
|
|
It uses perl.
|
|
|
|
|
|
|
|
Update to 2822t1f.
|
|
|
|
|
|
Fix changelog for dpkg 1.18.11 to mention PIE got enabled by default for
all architectures, not just the ones where gcc does that itself.
When emitting PIE flags on architectures where gcc does not inject those
itself, do it via a specs file too, so that maintainers can use them
unconditionally regardless of the object being compiled or linked.
When injecting -no-pie for linking via gcc specs also inject -fno-PIE.
Update the documentation to make the current situation more clear.
|
|
Pin the timezone to UTC so that the computed dates are deterministic
and produce deterministic generated man pages.
Closes: #843469
Signed-off-by: Guillem Jover <guillem@debian.org>
|
|
Missed in commit 10264ea2e5dc7b5df31bffae6fd644638d89ee15.
|
|
Commit 9d7ba99cc3ff84fc553ed39da9d2e4f4008d35b6 renamed the
triplettable file to tupletable and added the necessary dpkg
dependency to libdpkg-perl, but did not forbid the other now broken
combination of old libdpkg-perl and new dpkg.
Closes: #843429
Signed-off-by: Guillem Jover <guillem@debian.org>
|
|
Closes: #843436
|
|
Non-Linux systems allow to open directories, which return their
contents, this makes the load method fail with parsing errors. Make
the code only try to ever load regular files.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The scripts use modules not present in perl-base, so to be entirely
correct we require the perl package, even though it gets pulled anyway
by the libdpkg-perl package.
Warned-by: dpkg-gencontrol
|
|
Add support for compiler built-in features, so that we do not set
them when enabled and set negated flags when disabled.
We use gcc spec files to set these flags so that we avoid any conflict
with other incompatible flags that would make the build fail.
Closes: #835149
Based-on-patch-by: Bálint Réczey <balint@balintreczey.hu>
|
|
The get_timepiece methods might return an undef value if there was no
timestamp in the trailer. Do not call the Time::Piece epoch method in
that case, as that causes a perl error.
Regression introduced in commit ea22158eb8b0dcaf42b0cdacd5b4560764f353c8.
Closes: #843248
Based-on-patch-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
|
|
Trying to rmdir(2) or unlink(2) a non-existent pathname on at least
Linux returns with EROFS. Handle this case specifically to check if
the pathname exists with access(2).
Closes: #838877
|
|
The Debian build daemons have already been enabling parallel builds for
a some time now. This has the advantage that it should make the builds
faster, and that it gives similar conditions to local builders to the
ones on build daemon networks, helping to catch problems with parallel
build support in packages before the upload.
Closes: #842845
Suggested-by: Simon McVittie <smcv@debian.org>
|
|
Honor parallel build option.
|