Age | Commit message (Collapse) | Author |
|
In gdb/binutils, we can have random build problems if the
timestamps of intl/plural.[yc] are not in the expected order.
Quick hack: make sure plural.c is more recent than plural.y
such that we do not attempt to rebuild it.
Change-Id: Ieb5c543bf8aec313475fbbb76b4551d5d3d67019
|
|
problem with jenkins.sh"
This reverts commit 746a2d3e4f70ee300efae2b85237b5874ccbd85b.
This patch is causing all other validations to fail.
Change-Id: Iaa6cd52588663b573ab1ab055825a5c0cca4cf3c
|
|
jenkins.sh
Change-Id: Idf35c0ab734bf5ec049d26691f5f7b17ea28e881
|
|
Some build systems have race conditions (e.g. GCC
target libraries with multilibs), so it's safer
to run make install without parallelism.
See GCC PR 42980.
Change-Id: Iefbcd620473a771a69b82e7135cf3e8fd3723a84
|
|
Change-Id: I353c4aee38f3f350eb6c3a9b726db921dc2a7dc4
|
|
Abe should not use the -i (--ignore-errors) option for make. This causes
errors which occur during building to be undetected.
This patch also removes -k (--keep-going) as it makes it harder to diagnose
errors from build logs.
Since "make check" is expected to have some failures, we don't want to abort
those when failures occur, so "-i -k" is retained for make check.
Change-Id: Ie23640c752968f052adfdcc92a80bba0e35c8619
|
|
Change-Id: If6f4069e3c2f98b7ba6a2e8210c81b91368ed58f
|
|
The component_init function can return errors, but they are not propagated
at the call sites.
Also move build_failure, build_success and test_success to common.sh so that
they are available earlier in abe.sh.
Change-Id: Idc9b2e59a2a3bdbffaf53598b1eb0cd36201871b
|
|
it's dash instead.
Change-Id: I2b819d18b52e35045684948084b125c35d4d43a6
|
|
Change-Id: I174abf69889309a3d576da7837b0742df342ae36
|
|
Change-Id: I0fe883d550934b17585f075df16de1fc01dc6120
|
|
Pass --progress=dot:giga when abe is run with output directed to a file.
Change-Id: I5c0ecd03571815b4e42f94ebe5e348a4b0f64f3a
|
|
Change-Id: I69310152fcb7393a9c46ad85095f1af1eda6503f
|
|
|
|
https://bugs.linaro.org/show_bug.cgi?id=2124"
|
|
|
|
|
|
Change-Id: I4edc49c30e8deeefd72c97916ee533b1aae5dbf9
|
|
https://bugs.linaro.org/show_bug.cgi?id=2124
Change-Id: Ibe223aedff90c75d472df4a7e7dd2e1ee4f69140
|
|
Change-Id: I19f174f3f56ff7068ca6c426fc63c04eaec054a9
|
|
When building in jenkins, jenkins.sh adds a build_url line to the manifest.
An example such line is:
build_url=https://ci.linaro.org/jenkins/job/tcwg-bkk16-buildfarm/label=tcwg-x86_64-cam,target=native/115/
As the url includes the string 'target=', it causes the manifest parser to
incorporate parts of the url in the target name, causing a cascade of error
messages. There is more work to be done to improve the robustness of
import_manifest() but this patch addresses the immediate problem.
Change-Id: Ie14d555b8f4d2ad607ed8c9ed0cc0babe6c6a32d
|
|
Change-Id: I58fb26e84b2e3903c284576d9c137a277df9b325
|
|
Change-Id: I8ed46668962711c4aa051fadb25c78f1e2e646cc
|
|
Change-Id: Iad2016cb3692c9320555a982311d5792b1a95ac0
|
|
|
|
Change-Id: Id323e27e378dd1626b6a8fd4513e57996e75948f
|
|
Change-Id: I03be5ada35feceeec7318b0b73514f9738809736
|
|
Change-Id: I8f943ed9afe6441a033f6de298d6f049dcad4cba
|
|
This reverts commit 770cd2a469b0421334427001e956336143439e50.
Change-Id: I33ee342b90c43c242b8d30aeb23c9c15c806fea2
|
|
There is an ongoing problem which means that expect is unable to spawn
processes. Detect the condition before running make check, and try to output
a little bit of debug info.
Change-Id: I3ab113e0c98eb31a82720f1f55af225a8a59ec49
|
|
This will avoid non-deterministic build behavior and will fix _Pragma3.c test.
At the moment gcc_update will run only when GCC is checked out from VC repo,
and we need it to also run for building from tarballs.
Change-Id: I04268bacff5ba5f6a647e4bcfd7c79b606e87067
|
|
dereferenced.
Change-Id: I20580163d3693921556c5f2333af67678a590162
|
|
Run contrib/gcc_update after checkout, so that timestamps are
as expected, to avoid false failures in the testsuite.
Change-Id: I5691b0e57b11edb0a2831bef50cad55bcd8c9ace
|
|
code path.
This is required due to the following scenario: A git reference dir
is populated with a git clone on day X. On day Y a developer removes
a branch and then replaces the same branch with a new branch of the
same name. On day Z ABE is executed against the reference dir copy
and the git pull fails due to error: 'refs/remotes/origin/<branch>'
exists; cannot create 'refs/remotes/origin/<branch>'. You have to
remove the stale branches before pulling the new ones.
Change-Id: I436b4961deb5496e3b28567bad59e22f91ee7d09
|
|
Fix for https://bugs.linaro.org/show_bug.cgi?id=1717
ABE is to be consistent with Ubuntu library layouts because there is
historic precedent for this layout in Linaro binary toolchains.
"slibdir" stands for shared-objects libdir. Therefore all shared
objects should go into slibdir. On Ubuntu systems this is /lib.
"libdir" is where static libraries and linkers scripts are to be
installed. On Ubuntu systems this is /usr/lib. "rtlddir" is where
the dynamic-linker is to be installed. On Ubuntu systems this is
/lib.
Change-Id: I19242cb2ce118505a7f53224e1956b738ad2cd85
|
|
Change-Id: I11fc57997f48a8a792a227348979a95d2d320fcd
|
|
Win32 installers will need to be built manually until
InstallJammer dynamic configuration is enabled.
Change-Id: Ib04a9d550da7149b39c2a692506c88c4ec23abc8
|
|
Change-Id: I6917d9b4a2552507865260bd002212296532a8be
|
|
Change-Id: Idaca0e2bd07a1865d8eceb1bf26b60c4cf39c0ab
|
|
Set using configparams instead of a command line option.
Fix for https://bugs.linaro.org/show_bug.cgi?id=1717
Change-Id: Ia73c2a29eb3a6ac08dd46e546b4ac64ff32e6122
|
|
|
|
[https://bugs.linaro.org/show_bug.cgi?id=1751]
This patch deliberately checks ${branch} and sets it to HEAD if it is
empty (only for checkout()) so that we deliberately checkout HEAD,
rather than by happenstance. Otherwise we get git 'fatal' warnings
from a malformed checkout which pollute the log and confuse users.
Change-Id: Ic21f5c7d61d37b9d5be3ae887c98029ed76378de
|
|
debugging ABE."
|
|
|
|
After recent improvements to checkout code fetch_md5sums started
to treat $git_reference_dir/md5sums as the intended source of md5sums
data. While copying of [mostly infrastructure/] files from
$git_referece_dir is OK and beneficial, md5sums is not versioned in
the filename, and appearance of new version on $fileserver does not
invalidate the copy in $git_reference_dir.
The patch has a double fix for the problem. Firstly, it removes code
from fetch_md5sums that handles $git_reference_dir, leaving two options
for md5sums: either be present in the first place, or be fetched via http.
Secondly, it updates scripts/update-snapshots-ref.sh to remove md5sums
after generation of reference snapshots.
Change-Id: Iaef5f9f729aef75f2b5ad0c60cf666fafd5d6f57
|
|
[https://bugs.linaro.org/show_bug.cgi?id=1751]
This causes ABE to abort early if inputs are malformed, and saves us
from unnecessary execution in such an event. This patch also adds
error messages to the get_git_<foo> services if they're returning
'1'.
Change-Id: I4fe7afc3d3efce3936f0c859dea6dcff7e2ba37f
|
|
Change-Id: I355909aea040f4cda1d7f6ffe3c915e70e3044fe
|
|
All release candidates will include -rcX on the end. Some release
candidates will include a spin number preceding the -rcX. Prior to
this patch the release string was generated incorrectly:
gcc-linaro-5.1-2015.08-rc1 => gcc-linarorc1-5.1-2015.08-rc1
gcc-linaro-5.1-2015.08-2-rc => gcc-linarorc1-5.1-2015.08-2-rc1
After this patch the 'rc1' is in the correct place.
Change-Id: I7e2084e1fc5ccdf93eacb345ed249f3cb20f57d2
|
|
The staging.git.linaro.org server will now 404 so we need to direct
all uses to git.linaro.org. The testsuite/test.sh now passes clean.
Change-Id: I952385b7893413ce363de818d1d060fd98c39da3
|
|
toolchain packages."
|