Hans Rosenfeld <rosenfeld%grumpf.hope-2000.org@localhost> writes: > On Fri, Feb 17, 2012 at 09:54:17AM -0500, Greg Troxel wrote: >> >> "Hans Rosenfeld" <hans%netbsd.org@localhost> writes: >> >> > Module Name: pkgsrc >> > Committed By: hans >> > Date: Fri Feb 17 13:54:02 UTC 2012 >> > >> > Modified Files: >> > pkgsrc/devel/scmgit-base: Makefile >> > >> > Log Message: >> > Fix build on SunOS. >> >> Can you explain why this is needed (in this case, and in general, in the >> commit messages)? > > Yes, of course I can. Obviously it fails to link without it. > > Normally this should be handled by the BUILDLINK_LDADD.iconv variable in > converters/libiconv/buildlink3.mk, but that's been broken for years and > even after many hours trying to debug this I have no idea why. So I'll > just use the same fix that's been in other packages for ages. > >> While I can more or less see that this change >> shouldn't affect other than SunOS (presumably Solaris only these days), >> I don't follow why this is needed, and if so why it isn't an upstream >> bug that should be reported there and have a tracker URL (ok, this is >> git, a mailinglist post) in a comment. > > It builds without problems outside of pkgsrc, so I don't really see why > I should report it. > Seems I mixed something up here, thats completely unrelated since we > were talking about libintl and not libiconv. Sorry for any confusion > caused. So, it seems this is workaround for a bug in pkgsrc. In that case, I'd like to see a PR filed about the bug, and a comment referencing the PR in every workaround, much like we ask that patch files have comments and links to upstream bugtracker entries. Putting in non-understood workarounds without comments leads to accumulation of lore and new people trying to understand pkgsrc reaching the wrong conclusion. Also, I think the commit messages should indicate that a workaround was added and reference the PR. "Fix build on FooOS" as a commit message does not lead to the reader understanding the change (since it doesn't actually summarize the change, but rather describes one effect of the change), and I don't how anyone could make a good decision as to whether to go read the diff or not. In this case, I wonder how the git build on NetBSD gets libintl (probably it is just looking in base, since I think bl3 doesn't hide that), and if the issue is that git should explicitly depend on devel/gettext-lib.
Attachment:
pgpaTGPekP4xn.pgp
Description: PGP signature