Summary: | Mk/Scripts/do-patch.sh: not failing on missing patch file | ||||||
---|---|---|---|---|---|---|---|
Product: | Ports & Packages | Reporter: | John Hein <jcfyecrayz> | ||||
Component: | Individual Port(s) | Assignee: | freebsd-ports-bugs (Nobody) <ports-bugs> | ||||
Status: | Closed Not Accepted | ||||||
Severity: | Affects Many People | CC: | portmgr, tatsuki_makino | ||||
Priority: | --- | ||||||
Version: | Latest | ||||||
Hardware: | Any | ||||||
OS: | Any | ||||||
See Also: | https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=250723 | ||||||
Attachments: |
|
Description
John Hein
2020-10-28 17:43:07 UTC
(In reply to John Hein from comment #0) Correction: I think this was broken in do-patch.sh from day 1. r533459 just shuffled things around as far as the particular piece that is addressed in this bug. In which case is this supposed to happen? If you have a use case, would a simple "set pipefail" at the top of the script not be enough? I have opened review D27007 and bug #250723 with a more global pipefail patch. (In reply to Mathieu Arnold from comment #3) See bug 250530 comment 10 for a real failure case (devel/llvm10 before ports r553456). pipefail seems like it is a good alternative fix, yes. I have not tested your wider pipefail patch from bug 250723 to see if anything breaks that shouldn't, but conceptually it seems like it is reasonable. It's not written in porter's handbook, but we can use make -D PATCH_DEBUG patch to debug the patch. Why don't you just run it before you ask them to commit? :) Closing this bug in favor of the fix in bug 250723. |