Misc. operability tweaks for clang-format
#5422
Conversation
Co-authored-by: Theodore Tsirpanis <teo@tsirpanis.gr>
johnkerl
force-pushed
the
kerl/clang-format-version-info
branch
from
afe70c8 to
d2b951b
Compare
|
|
||
| src=$GITHUB_WORKSPACE | ||
| src=$(dirname $0)/../.. |
There was a problem hiding this comment.
It turns out CI invokes this with source not with bash, and when you source a script, its $0 is -bash
There was a problem hiding this comment.
$ cat foo
echo '$0' is $0
echo '$*' is $*
$ bash ./foo 1 2 3
$0 is ./foo
$* is 1 2 3
$ bash $(pwd)/foo 1 2 3
$0 is /Users/kerl/git/TileDB-Inc/TileDB/foo
$* is 1 2 3
$ source foo 1 2 3
$0 is -bash
$* is 1 2 3
$ source $(pwd)/foo 1 2 3
$0 is -bash
$* is 1 2 3
There was a problem hiding this comment.
Maybe we should fix the workflow, it doesn't make much sense to run this script by sourcing it.
There was a problem hiding this comment.
🤷
I saw somewhen/somewhere some comments about sourcing in GH YAML; maybe a recommended best practice? 🤔
There was a problem hiding this comment.
Personally I'm happy with trying the experimental method -- I can modify the calling YAML to bash the script than source the script ...
There was a problem hiding this comment.
No idea. 😅
My intuition is that sourcing should be done only in special circumstances and when a script is less of a "function" and more of an "environment initializer".
There was a problem hiding this comment.
@teo-tsirpanis oh I 100% agree, and have for years, even before TileDB. The only good reason for source to me is something that needs to set environment variables, aliases, etc. for the parent process.
I'm simply wondering if there's some other good reason, within this particular GHA context, which the person who keystroked source into GHA YAML didn't bother to explain at the time, leaving you & me to wonder. But again, I'm happy to find that out by experiment.
There was a problem hiding this comment.
For whatever the reason (good or bad) whenever this was written, there's a lot of it:
kerl@arvad[prod][kerl/current-domain-error-message][TileDB]$ rg ' source ' .github/workflows/*ml
.github/workflows/unit-test-runs.yml
88: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/test-cloud-e2e.yml
57: source $GITHUB_WORKSPACE/scripts/ci/bootstrap_libtiledb.sh
68: source $GITHUB_WORKSPACE/scripts/ci/build_libtiledb.sh
106: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/release.yml
136: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/check-formatting.yml
41: source $GITHUB_WORKSPACE/scripts/ci/check_formatting_linux.sh
.github/workflows/build-ubuntu20.04-backwards-compatibility.yml
41: source $GITHUB_WORKSPACE/scripts/ci/bootstrap_libtiledb.sh
42: source $GITHUB_WORKSPACE/scripts/ci/build_libtiledb.sh
122: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/ci-linux_mac.yml
157: source $GITHUB_WORKSPACE/scripts/ci/bootstrap_libtiledb.sh
168: source $GITHUB_WORKSPACE/scripts/ci/build_libtiledb.sh
189: source $GITHUB_WORKSPACE/scripts/ci/posix/build-services-start.sh
211: source $GITHUB_WORKSPACE/scripts/ci/posix/build-services-stop.sh
221: # It relies on finding its source directory.
227: source $GITHUB_WORKSPACE/scripts/ci/build_benchmarks.sh
274: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/build-rtools40.yml
66: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
.github/workflows/build-windows.yml
432: source $GITHUB_WORKSPACE/scripts/ci/print_logs.sh
|
The |
|
@teo-tsirpanis please feel free to merge when you are ready. I lack permssions on this repo. Thank you! |
|
Thank you @teo-tsirpanis ! |
Found while debugging CI fails on #5421.
TYPE: NO_HISTORY