Summary: | net-p2p/syncthing-1.29.5: fails tests (-buildmode=pie not supported when -race is enabled on linux/amd64) | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Sam James <sam> |
Component: | Current packages | Assignee: | Marc Schiffbauer <mschiff> |
Status: | CONFIRMED --- | ||
Severity: | normal | CC: | fkobi |
Priority: | Normal | Keywords: | PullRequest, TESTFAILURE |
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: | https://github.com/gentoo/gentoo/pull/41941 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | build.log.xz |
Description
Sam James
![]() ![]() ![]() ![]() The patch was dropped in: commit 381cfa494305a29eed29c54255d4395dbfc66ab0 Author: Filip Kobierski <fkobi@pm.me> Date: Thu Dec 26 20:31:09 2024 +0100 net-p2p/syncthing: add 1.29.5 with improved ebuild changes, from top to bottom: - do not rename distfile - move project primary license to be the first - made RDEPEND more readable - updated DOCS - removed race patch - added loops to shorten the ebuild in src_{prepare,install} - swap $(usex a b "") to $(usev a b) - use brace expansion instead of for loop for installing tools - swap "syncthing" to ${PN} where it makes sense Signed-off-by: Filip Kobierski <fkobi@pm.me> Part-of: https://github.com/gentoo/gentoo/pull/39851 Closes: https://github.com/gentoo/gentoo/pull/39851 Signed-off-by: Marc Schiffbauer <mschiff@gentoo.org> It doesn't say *why*, though (please say that in future -- if you dropped it because it doesn't apply anymore and you don't see why it was there, say that in the commit message). At https://github.com/gentoo/gentoo/pull/39851#issue-2760089746, it says: > @mschiff could you please explain what is the tests_race patch for? > The patch does not apply as upstream changed the build script. > Do we still need it? FWIW, the patch was added in: commit 258b078422e2aad516119787da45a8c2d8c0aa78 Author: Marek Szuba <marecki@gentoo.org> Date: Wed Mar 8 11:32:23 2023 +0000 net-p2p/syncthing: add 1.23.2, drop 1.23.0 Signed-off-by: Marek Szuba <marecki@gentoo.org> I'm also surprised that neither of you hit it, because it should happen (but maybe it doesn't) with PIE, which we do in: go-module_src_unpack() { if use amd64 || use arm || use arm64 || ( use ppc64 && [[ $(tc-endian) == "little" ]] ) || use s390 || use x86; then GOFLAGS="-buildmode=pie ${GOFLAGS}" fi Created attachment 927695 [details]
build.log.xz
I see that in the build.go "-race" is mentioned twice.
Since we target PIE, which conflicts with that flag, maybe it would be better to have a sed 's/-race//g' instead of the patch?
> > Do we still need it?
I was hoping for an answer before it gets merged.
Also Sam, could you please edit the title to contain both flags?
Like "${P} fails tests: -buildmode=pie conflicts with -race".
(In reply to Filip Kobierski from comment #4) > I see that in the build.go "-race" is mentioned twice. I'm still curious as to how you can't hit it -- do tests really pass for you? > Since we target PIE, which conflicts with that flag, maybe it would be > better to have a sed 's/-race//g' instead of the patch? It's fine with me as long as people use qa-sed to notice if it stops applying. > > > > Do we still need it? > I was hoping for an answer before it gets merged. I don't blame you ;) > > Also Sam, could you please edit the title to contain both flags? > Like "${P} fails tests: -buildmode=pie conflicts with -race". Sure. |