Age | Commit message (Collapse) | Author |
|
Change-Id: I9b750054b5f9ab5c60c9fb06a57b849805781f4b
|
|
Ideally we want profiledbootstrap for each bootstrap config, but I am
not sure if that's supported in practice. Thus in the patch, profiledbootstrap
is enabled only for normal and LTO which are known to work.
Change-Id: I0652a8ee4b64f1e9a66d59481bea52eb21d18a99
|
|
Change-Id: I53fc3879608ee9654ba2b8b7da5b86c9b7099269
|
|
Change-Id: I21fad361d7b12c9a15617069073daf44ca56894b
|
|
tcwg_kernel-build.sh.
Change-Id: I4c47f486d78f6e9a20732e06a387bb0db77c59ed
|
|
Change-Id: Ibc6dcd085f684eb6f09ec9a512aa5abd6b9d8200
|
|
... these are no longer used after refactor of tcwg-update-snapshots_ref job
Change-Id: I07e16c2c60e00cc629e5eef9fcd6798aaab50cc3
|
|
Turns out that condition in "if ${undefined_var+false}; then ... fi"
evaluated to "true" when $undefined_var is undefined -- i.e., empty
condition is a "true" condition, as long as it is not physically
empty ("if ; then ... fi" will produce a syntatic error).
This perculiarity of bash caused unexpected behavior and we used
"empty" branch of base-artifacts for several builds -- which,
unfortunately, caused us to loose history regressions in several
configurations. We caught this quickly, so no serious damage done.
Fix by adding verbose conditionals.
Change-Id: Ib6f1a624a9ea41f7223bc3e85e2e4eddc3fad9a9
|
|
... which no longer holds. After we started keeping first-bad
artifacts (which are very useful), we can have no change in result
metric compared to previous [first-bad] build.
Change-Id: I9a3651226ca12c20ae4c06a2b2bd00bd8a63f518
|
|
"$@" expands to several strings, which become extra git command-line
arguments, and "$*" expands to a single string.
Change-Id: I95799d258ffb8dc3e4f79777894a0425eafeceb9
|
|
... to remove old results and avoid filling up the disk.
Change-Id: Ife87756fc92f48717baa0f84c7c3fe16b49b1899
|
|
Change-Id: I946de0b6d4d4ee62f1bae1532ad9bff52131bcb5
|
|
In case of last_good being bad we often have first_bad just on
the boundary of bisection range, and last_good outside of bisection
range, so we end up ignoring last_good commit in bisections of similar
configurations. Adding first_bad commit will reduce bisection time in
such cases.
Change-Id: I33ab875ec751a161452634c8b989a39b4e8eb05a
|
|
Our new tcwg-tk1-10 board has tcwg-tk1_32-build only.
Change-Id: Ifaa68fbadc83f6e1b0cca0cdf520587ef210d77c
|
|
When re-triggering bisections due to bad $last_good commit, we
need to check that baseline_rev is a parent of last_good -- otherwise
we will end up with invalid bisection interval.
Change-Id: Ic9f9e07783c967f2aef777105a26f02e526a8505
|
|
Change-Id: I4d88a4ed947c2bd6f23f216957dc1a141a29cf5b
|
|
Return an empty list instead of an error in case the label contains no
slave.
Change-Id: I5b1561dd86d62347b1db59b3204c1c955d181000
|
|
Change-Id: Ibd291b23a3c4228f548cdf270fb5cb719b7033b5
|
|
Change-Id: I4c5ccf6fed723c2661cefb5a650eb19e5be0667b
|
|
Add a few more messages, to make the Jenkins console easier to understand.
Change-Id: I2bab8aaf5333c4e7f8d8d8681986c961336f1fd6
|
|
process to kill.
Handle two cases where killall can fail:
- command not found, in which case we want an error so that we know we have to install killall
- no ssh-agent process running, which is OK
Change-Id: I0d6d7e40eed6db7c6250f2231f05b8cca73c8355
|
|
Change-Id: I6bb8e3121ebae9af22c4b9b0593d7a35a2ad35fb
|
|
In print_tester_label_for_target(), do not use tk1/tx1 anymore, use
APMs instead.
However, this means we no longer use armv7 hardware to test arm-*
configurations (as opposed to armv8l*).
Change-Id: Iff08678c39ad15f59628e5a91f62d0812d0bb0b3
|
|
Change-Id: Ie3b51996e1ccc56ebd2c6bfb41982c950a7ee340
|
|
... since we no longer use it in check_regression().
Change-Id: I2014125ee0f17a2ddc43603a626efc69177f4709
|
|
Change-Id: I1b025757b2f4b1f845b75b88eac5004f3a186a92
|
|
Change-Id: I00df2aaa065377884b60360060ed2234e6555e3a
|
|
Instead of a hack in docker-run.sh add ability to pass environment
variables in remote_exec(). Use this to pass ssh_host/ssh_port via
container_exec() function.
Change-Id: Ic8f7bb3c3ae94b501eb475e90b892c0d769dd0dd
|
|
... which will give path with @build_num@ resolved.
Note that setting results_id=... in benchmark() doesn't override
the global value of results_id.
Change-Id: I0708db2b56d1683845050c6d5fd9bb4f5cd2933d
|
|
We have disabled ssh connection sharing for bkp-01.tcwglab, so
there won't be "5min hang" problems and it has big disks.
Change-Id: I3f77217b01de7b50a72feaa1680e09c8127de076
|
|
processes.
Without this patch, failure to kill such processes is not reported as
part of the script exit status. This happens for instance when the killall
command is not installed on the machine.
Change-Id: Ic0fa10ce6b69f12fb2382cfa3680fcc1c7d1b4bd
|
|
Specify binutils_branch and gcc_branch for gnu builds.
Change-Id: I52cd3d0d2837f1fe3c5c492dd657cc6c71408100
|
|
Rename bad_rev and baseline_rev to bad_branch and baseline_branch,
and resolve them to SHA1 revisions.
The motivation for this change is ability to gather historical data,
e.g., code size regressions from GCC/LLVM 5 to 6, to 7, to 8, and so on.
The bisects are intended to be triggered manually, which is why
branch handling is desired.
Change-Id: I9c46270e0092e72836986a4bed08799ec6d84a00
|
|
Use --manifest value when it's provided.
Change-Id: I7c4667c6c607e14ee31eafe2f0fee958abb962cb
|
|
... for LTO
Change-Id: I2fd43c6216fe1ae57ba656da4a0e101fa58df471
|
|
Change-Id: I40b52cebea7a92a4c04bb7c7a044d0f36a95e19f
|
|
Change-Id: I90a53779fd98af8338de56a80ffad416bbe1b14a
|
|
... otherwise clang tries to use x86_64-linux-gnu target.
Change-Id: I2583fde555ad08afe0f4a76905147ff7f8055c23
|
|
"<build_num>" gets escaped into "\<build_num\>", which is not matched
by "sed" in tcwg-benchmark.sh. Use "@" to avoid escape characters.
Change-Id: I7e76975cc8561dbd1eeb1caa48dabaada09d6312
|
|
If manifest.txt was uploaded in job's parameters, then
pass --manifest option to MakeRelease.job.
Fix MakeRelease.job to not pass --target when manifest is set,
otherwise ABE complains.
Change-Id: I2aa25d3840281159e2ed00cfaab3b9e47a661801
|
|
Change-Id: I0885d8c1f7d228cfb5abaa5dd986a6c0d07dff24
|
|
"Echo" concatenates all elements into a single string, so we
need to break them up for "grep" to match its pattern.
Change-Id: I954ca80fc6f0e289185780f76988688ce0f1ed78
|
|
Many configurations seems to fail. Below are some examples:
- llvm-master-aarch64-lts-allmodconfig -- produces 2k less .o files
- llvm-master-arm-next-allmodconfig -- produces 4k less .o files
- llvm-master-aarch64-next-allmodconfig -- produces 5k less .o files
- llvm-release-aarch64-lts-allnoconfig -- can't boot
- llvm-release-aarch64-mainline-allnoconfig -- can't boot
Change-Id: I5e88e83579e6acbf75aaf850b3defcf5d4ed850a
|
|
It is expected that developers using clang to build the linux kernel
will also use LLD for the linker. We would like to test this in the LLVM
builds of the linux kernel as much as possible. To use LLD in the kernel
build we set LD=/path/to/linker when calling make.
At the moment we use LLD for llvm-*-aarch64-*-* builds and for
llvm-master-arm-*-* builds.
Change-Id: Ib1d543480278abec193cde6aeba0f457bcd603aa
|
|
We use the same form as clang, copying lld from the monorepo into
tools/lld.
Change-Id: Ieff55d1aafe2a49c87f3c7e518933a5adebcf8ff
|
|
Change-Id: I969944e551ffce03e1e8cb19f205e3288b60d4e8
|
|
Change-Id: I7872776502e64b48db822558e5ea4171afd38e72
|
|
Since $BUILD_NUMBER can come from either build or bisect job,
it's not that informative, so just not use it. Also, it is not
clear what to specify for BUILD_NUMBER is manual runs.
Change-Id: I6cdbc17af7efc5a362452c55461998e85f723229
|
|
Change-Id: I74c3884cdbf0fed39dcdd51411effdc935d23545
|
|
I missed this change in the rename commit.
Change-Id: I82e5e93fe52e8e53d0ca9cfc95b70dda8485e2fb
|