William Hubbs posted on Thu, 13 Dec 2012 21:51:33 -0600 as excerpted: > On Thu, Dec 13, 2012 at 10:06:34PM -0500, Ian Stakenvicius wrote: >> -----BEGIN PGP SIGNED MESSAGE----- >> Hash: SHA256 >> >> On 13/12/12 06:49 PM, William Hubbs wrote: >> > For example, glibc-2.9 and gcc-2.95. I think that if we are going to >> > keep things this old in the tree we need a good reason for them. >> >> iirc, gcc-2.95 and linux-2.4 (still used for some embedded systems) >> play best together. > > I'm not sure how strong this argument is because we don't have any 2.4 > kernels in the tree, and I am wondering why we still have a > linux-headers-2.4.
This only relatively indirectly affects me as my focus is amd64, where even gcc-3 has been out of support for quite some time, but from what I've read, there are still people who find old gcc-2.95 good for building old code or for various other special projects. And given its wide distribution due to Red Hat, there's probably a lot of old code that requires 2.95. As such, I'd suggest keeping it in the tree as long as someone can be found to keep it building at all on current systems. Once there's nobody and it goes horribly broken so it's of no use in any case, then yeah, remove it, but unless/until that becomes the case, gcc-2.95 really is a bit of a special case, and probably does still have a few users supporting their strange corner-cases. I'm no gcc history expert, but if I'm not mistaken, 3.1 has a similarly wide historic distribution (next redhat version?), and it's likely there's a reasonable case for keeping one of each gcc minor, presumably the last, in the tree. However, that still leaves several as cleanable, with the oldest being 3.2.2. Or it may be that 3.2 and 3.3 can be removed as well, leaving 2.95.3-r10, 3.1.1-r2, and 3.4.6-r2, as the only pre-4.0 gccs. 4.0 might be removable too as a non-major-distribution early 4.x, leaving 4.1.2 as the first 4.x gcc. Obviously there's folks way more equipped to make those decisions than I, however. I'm simply pointing out that I know for a fact that 2.95 has some serious history behind it, such that there really are quite possibly people using it for something or other, even if it's nothing even close to their primary system compiler, and that gcc in general is arguably a rather special case, with good reasons to keep quite a few more versions of it around than for most packages. It may be that linux-headers-2.4 is a similarly special case, even with no 2.4 kernels remaining in-tree, because if there's any single package people are likely to procure and build independently of gentoo, it's likely to be the kernel, and there's a significantly higher than zero chance that someone, somewhere, is still building their own 2.4 kernels and using the 2.4 headers, not necessarily for the gentoo build system, but very possibly for some embedded system or other special project they have going on. glibc would of course be in the same rather limited family, where a gentoo build system and toolchain packages are being used to support unusual corner-case projects. I'm not sure about anything else, as the build host's PM and support packages could be current gentoo mainline, just the target some weird corner-case, but again, toolchain folks would know more about such requirements. -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman
