Age | Commit message (Collapse) | Author |
|
in jenkins-helper.sh for now as it will be used by chromium_build and bisect scripts
Change-Id: I249234f14b077e4ecfc771ea01812ab53dce73aa
|
|
After a recent change in validate_failures.py, the notification reports now include a few undesirable lines like:
# of stable fails 61
# of flaky entries 9
# of all tests 890731
which are irrelevant in the notification: we want to keep only the list of regressions.
This patch removes all lines after "Results Summary" (included).
Change-Id: I192d495161817e5faa6a63408cb2c323d193ff84
|
|
Change-Id: I8748e1d0a5440c09827c223379690fabc5dae0b9
|
|
Change-Id: I8c71b573ed0f7502c8156032aa07e03f031a0388
|
|
Change-Id: I9a2f5a68ad69dc9188dad4638ced528cda2a19c3
|
|
workspace git dir
Change-Id: I86f5b69f94c66e45223652e30e526f1f12356e09
|
|
The LNT sample_variability datas will now contain many inputs that we can
display separately :
- <bench> (avg) (percentage) : the noise computed by avg algo (percents)
- <bench> (avg) : the noise computed by avg algo (absolute values)
- <bench> (max) (percentage) : the noise computed by max algo (percents)
- <bench> (max) : the noise computed by max algo (absolute values)
- <bench> : The one used in the algorithm for regression detection (=avg)
Change-Id: I9e770c10b462a2e0e587af88ff54ac9e5253b27f
|
|
Change-Id: I1ee803ffc57eab2afa4e4aeb93b137ec374f0969
|
|
using avg
.. But the noise displayed on the lnt dashboard is still max / 2-peaks-linear algorithm
Change-Id: Ia45013c3deb2bbdc4876fd073541044153b580c6
|
|
When the bmk boards are unreachable, the tcwg-update-bmk-containers
project change the <board>.lock file into the <board>.bak file. which
makes the board unusable by the tcwg-benchmark projects.
Then, in that case, the tcwg-benchmark_backend project cannot start
because it cannot find any board to run on.
'Ran out of boards to try -- giving up.'.
The tcwg-benchmark.sh script will never try to reboot the board.
Meaning, the unreachable boards cannot come back into the pool.
Once disabled (<board>.bak), the boards are only checked by the
tcwg-update-bmk-containers project. It needs to be power-cycle if
unreachable.
Change-Id: Iaf0f399a43d292f917b3631ee7e7114f5abc8c1e
|
|
It turns out "git am", "git am -p1" and "git am -p0" all have
different behavior when applying patches and skipping
already-applied patches. So try all three approaches.
As an implementation detail, rework parameters of apply_series()
and apply_series_with_retry() to enable easier customization of
"git am" options.
Change-Id: I123373d551a6ff81e750558c0f2c91e04db6dff4
|
|
... configurations.
Change-Id: I3443c001ad1904da9a99dcd834860f6591939d09
|
|
to provide pointers to LNT health check pages for various projects.
Change-Id: I983a342f8ac0ae031a62ea399c0db328e6d53f4b
|
|
Change-Id: I4f50c4e80365260e8bedc9c406312deb88a311fa
|
|
Change-Id: Icbca7b0b805f9739f1a5d1db2c936e687d99ed9a
|
|
.. and define rr[components] dynamically
Change-Id: I076e1e6e2c486b8d8b3fe9c0754e40029e5b9037
|
|
Change-Id: I69867ef7eef504995b416e809743bb924f048a89
|
|
Change-Id: Ie545af4d86966909107b60ce4e757ccac609c924
|
|
Change-Id: Ic5d1d9e78cc93ccb4c06cee6a38111e7bcea21fd
|
|
Change-Id: Ic21b6ecaf7b05ba0215c3031271e39f7884c26c1
|
|
Change-Id: I2c1539f1dca1c46d4404b9880c8d6a73880da322
|
|
.. and problematic because it restore the git as initially
Change-Id: I8a841b7f6a23d6425bf1882c7f0cb1079e79e2c5
|
|
This is a first version, for the immediate need for chromium design.
A more advanced version could be useful for the CI. Will be redesign
later.
Change-Id: Ie635cc9c016f5e8e47c402fbd1d9b0ff8fab3242
|
|
.. BTW, sorry for the spams ;-)
Change-Id: I810308cea35d0f578d819a7d18b416518998f3ad
|
|
Change-Id: I9e32bc7552f189df28d9871a6b8cec1bca4d7502
|
|
so that its purpose is hopefully clearer
Change-Id: I6a65c4a63f195e67c014097d8f233b40d84e9f4f
|
|
installation
Change-Id: I900e20be8f24bcc25c298b276b1db4a6f949caf3
|
|
Change-Id: I7d7122762b3bf2f5df2ccd9e4057d5909de96f77
|
|
Change-Id: Icc9b463e2fc0a07f399925cbe59fd0a573f2ad52
|
|
either a branch or a sha1
Change-Id: Iaa85a70cba7aed2c9d1a012ac8bf8ec96bdde1c2
|
|
.. and skip component update if dep is unknown or directory unexisting
Change-Id: I548f3e4729715e298b2c7120c97808803363a8a9
|
|
Change-Id: I5162cc625ac32e7ea42af769c3d4dbe17972805e
|
|
Change-Id: I6d404919d281423e636fa92d47aa3614d8d252c6
|
|
Change-Id: If128cdf0ec1f0ba8cf29513fd0a0eb81dac68af3
|
|
Thus, the bmk-specific-variability-xxx.csv is expressed as a percentage.
Whereas the displayed variability handled by lnt is an absolute value.
Change-Id: I9d51b94ef670bc678525e365418cb22ac8c5c6ed
|
|
Change-Id: I8792ea1eef38e52c5159e02610198d6b6f24d13c
|
|
There have been some false positives in both GCC and GDB regression reports
because flaky entries expired before the CI could re-detect them. Address
this problem by increasing expiration date for flaky entries from 6 to 12
weeks, and by increasing the window where we try to detect flaky entries
before they expire from 4 weeks to 8 weeks.
Also, fix typo in one of the comments (s/what/that/).
Change-Id: Id3458a42591da4291054b4bdbeb6f64e63cd2838
|
|
Change-Id: I426b7f673262877761bf7af24bc5b8016b70a349
|
|
Change-Id: I05bc70d2cda02f8610eb3516f4afdad578565e10
|
|
I recently ran a manual tcwg_gdb_check--master-arm-precommit job, and set
the gdb_url parameter to an "https://git_url#branch" value. This meant
that when round-robin-notify.sh ran, there was no pw directory (even though
a --pw_dir argument was passed) and thus the $pw array was empty.
The result was that an email was mistakenly sent to the gdb-testers mailing
list, because print_mail_recipients() checked that the value of $pw[gdb]
was empty and thus thought that it was dealing with a postcommit build.
It can be argued that I should have used a build job rather than a
precommit job to test my git branch, but in case someone else has the same
brilliant idea this patch makes the script do the right thing in this
situation.
Change-Id: If8454a83b471d519446952558bffcb640a5063f8
|
|
Change-Id: Ic8f913733cca75b73c9492baf31ff0875cdb73cd
|
|
also generate a file for recent annex files
Change-Id: Ic4ff0109eaeaa1c105b8a7f1ec0628ed8e87bb59
|
|
cleanup
Change-Id: I5d9f4cf2d8d6c3d05a641c272dd67e846a6a88d5
|
|
Change-Id: I19fe9a370f1e1da9baddeedb1d157625fd3a80fa
|
|
This patch makes the logic to apply patches (and retry with different
-p level if needed) common to precommit-ssh-apply.sh and pw-apply.sh.
Change-Id: I65f0484c0087bfee79c569a9a49c69e049b9d0a7
|
|
Some contributors do not use git format-patch, and their patches do
not include the "a/" and "b/" path prefixes, making 'git am' fail,
since it defaults to using '-p1' to suppress that leading prefix.
If applying the patches with -p1 fails, this patch retries with -p0.
This will help us handle more patches in precommit CI.
Change-Id: I91ccc267c6ec1d08974f2530c98096c65f7b2fee
|
|
Change-Id: Id92f1e97d7e5032cdb5b91874119f11459f321d2
|
|
Change-Id: I78eb602ba96aedf82b4fa224ccabbfd87db84fda
|
|
output-bmk-results
Change-Id: I9a2cf890425cba9e5b022462911c2ae0f9ac188f
|
|
Change-Id: I17779d0ab850dcd4fb6041ac3dc3536671f09951
|