Summary: | Make run-jsc-stress-tests --remote work on Linux too | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Csaba Osztrogonác <ossy> | ||||
Component: | Tools / Tests | Assignee: | Csaba Osztrogonác <ossy> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | commit-queue, darin, ddkilzer, fpizlo, oliver, ossy | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | Unspecified | ||||||
OS: | Unspecified | ||||||
Bug Depends on: | |||||||
Bug Blocks: | 141165 | ||||||
Attachments: |
|
Description
Csaba Osztrogonác
2015-01-28 09:54:20 PST
Created attachment 245549 [details]
Patch
I have no idea if it will work on iOS, so it would be good if an iOS expert can review it.
Some additional info why we need this changes: - The runner script is run with "sh runner" command and sh is dash on Debian/Ubuntu distros, so we have to remove the function bashisms. - find doesn't have "-depth N" option on Linux, but "-maxdepth N" is available on Linux and Mac too and do the same in this case. - find warns if we put an option after a non-option argument, so we should put is before the -name. ("find: warning: you have specified the -maxdepth option after a non-option argument -name, but options are not positional (-maxdepth affects tests specified before it as well as those specified after it). Please specify options before other arguments.") - trap of dash doesn't know SIGINT SIGHUP SIGTERM signals, but bash and dash know INT HUP TERM signals, so let's replace them. Comment on attachment 245549 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=245549&action=review > Tools/Scripts/jsc-stress-test-helpers/shell-runner.sh:24 > +numProcs=`sysctl -n hw.activecpu 2>/dev/null` A lot of times this pattern is dangerous; if there is some failure it is silently sent to the bit bucket. Do we really need to do this? Comment on attachment 245549 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=245549&action=review >> Tools/Scripts/jsc-stress-test-helpers/shell-runner.sh:24 >> +numProcs=`sysctl -n hw.activecpu 2>/dev/null` > > A lot of times this pattern is dangerous; if there is some failure it is silently sent to the bit bucket. Do we really need to do this? It always dumps error on Linux: "sysctl: cannot stat /proc/sys/hw/activecpu: No such file or directory" We don't need useless error/warning messages always on Linux, and we already use the same pattern for nproc in line27 and in jsc-stress-tests too. I think it's safe, because we handle the return code/value properly everywhere. Comment on attachment 245549 [details] Patch Clearing flags on attachment: 245549 Committed r179343: <http://trac.webkit.org/changeset/179343> All reviewed patches have been landed. Closing bug. |