Re: Daily build compiler
Greg Troxel <[email protected]> Mon, 15 Jun 2026 10:50:48 -0400
| Newsgroups | gmane.os.netbsd.devel.packages |
|---|---|
| Message-ID | <rmijyrzzwl3.fsf__48463.57466174$1781535064$gmane$org@s1.lexort.com> |
Jonathan Perkin <[email protected]> writes: > I don't want to start another platform debate, but it's worth pointing > out that GCC 15.x is already quite common across other platforms that > pkgsrc claims to support, and e.g. Fedora 44 is already on GCC 16.1, > so it would be helpful to qualify such statements to "for *NetBSD* > branch users", as it's not true that these fixes aren't going to > improve the branch for *any* users. It's news to me that Fedora is on 16, and if 15 is common on platforms (releases of) with claimed support, then it's certainly fair to accomodate. I was operating under the impression, gathered from reading tech-pkg@ (where the community is centered) and commit messages, that gcc15 was in the pkgsrc world only turned-on-in-your-builds-to-see-how-it-was, where normally smartos used something something less new. Basically I see it as unhelpful to have a bleeding edge system compiler when upstreams have not caught up, unless a system values surfacing those issues over having things work (perhaps that's what Fedora is for). I'm still not ok with labeling things gnu89, absent upstream doc of that spec*, for packages that were building fine with slgihtly older compiler defaults, likely meaning c17, and if not c11, and certainly ok with c99. That deserves discussion, which has been missing. * which would IMHO make the package something that should be considered for deletion, as likely not being maintained by upstream for over 20 years. Not that we should delete because of that, but it makes the question worth asking.