On Thu, Oct 17, 2013 at 06:10:51PM +1030, Ron wrote: > On Wed, Oct 16, 2013 at 11:49:08PM +0100, Colin Watson wrote: > > Package: vpb-driver > > Version: 4.2.55-1 > > Severity: important > > Tags: patch > > User: [email protected] > > Usertags: arm64 > > > > vpb-driver's config.guess/config.sub are out of date for the forthcoming > > arm64 port. The attached patch sets things up so that you don't have to > > be bothered by this type of bug for future ports. > > Thanks for the patch, but you realise this shouldn't actually be a problem > in the version that's actually shipping in _Debian_, right? > > Unless I'm missing something, aarch64 support was added to autotools-dev > in Feb 2012, and vpb-driver already has new enough versions for that. > > > If that's not true, then please do let me know what other version of them > you do need for this. This patch just hides the problem for Debian, but > I'd much rather fix it for all users with a new upstream release whenever > this sort of thing is needed. Since then I'll only be bothered by people > reporting bugs against old versions they haven't updated, not people using > the latest version on systems other than Debian ;)
I was one version back from the latest version in Debian, indeed, and the latest version does have new enough files. However: We port to new architectures an average of once a year or so. Every time we do this, we have to fix a whole bunch (hundreds, at the very least) of packages shipping old config.guess and config.sub files; it's by far the most common porting problem by package count. It is very tedious work and means that the porters end up wasting time doing this rather than dealing with more interesting porting bugs. In quite a number of those cases, the maintainers of those packages have previously said that they'll just release new versions for this kind of thing. However, life happens: sometimes they've gone MIA for a while, or sometimes they have some external constraint that means they don't want to upload new versions just now. None of this ought to hold up the porting work. Every package that updates its config.guess/config.sub automatically at build time is one fewer package with a non-zero probability of being a hassle for *any* future port. I used to think the same way you do: hey, I'm a conscientious maintainer, *of course* I'll update config.guess/config.sub when I get bugs about it. Then I started doing ports to new architectures and saw things from the other side of the fence, and now I see why the autotools-dev maintainer has been banging on about this in README.Debian for all these years. There's an enormous time lag on getting config.* updates into packages by way of source uploads. We're in a situation where lots of otherwise-portable packages routinely fail to build during architecture bring-ups just because they don't recognise the architecture name yet and have to be manually re-educated, and this doesn't really seem a sensible situation to be in. > Thanks for the heads up, but please check the current version and let me > know if there's still actually a problem. I can get a new package out > quickly if there is, but afaics, we're already fine. Sorry if this bug isn't quite right as stated. However, statistically, I just filed this bug too early. :-) Cheers, -- Colin Watson [[email protected]] -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
