Bug#291631: cmp/diff/etc. lack PT_GNU_STACK header
On Mon, 24 Jan 2005, Jeroen van Wolffelaar wrote:
> On Mon, Jan 24, 2005 at 01:46:30AM +0100, Santiago Vila wrote:
> > That's the correct explanation, yes. It has never been a bug to build
> > a package using stable if the dependencies are compatible with the
> > ones in testing. In this case, Pre-Depends: libc6 (>= 2.2.4-4) is ok
> > because sarge has 2.3.2.ds1-20 and libc6 is not in oldlibs.
>
> Well, it is recommended to build in sid, also, I do think building in
> woody is bad for several reasons:
> - you do not ship what's in sarge, you cannot reproduce the build
> because it was build with software not in sarge. This is often the
> case in minor ways, but intentionally doing so seems awkward at best.
I agree it's a little bit awkward, but the build is done on 10
different architectures. If there were a FTBFS problem, chances are
that I would be notified about it.
> - only one architecture will have these older woody depends, all other
> architectures not. I don't know what reason you have to build on
> woody, but that's defeated by this buildd reason.
If somebody using the i386 architecture (which is the most common one)
reports a bug and it's fixed in the package in sarge, I can point the
user to the URL of the fixed package, and he/she can install it
without having to upgrade libc6. I think this is definitely a good thing,
but I now agree that it might not compensate for the potential bad things.
> - Suppose one day after sarge is released a security update needs to be
> made, and this package is suddenly build on sarge. There might be
> weird bugs hiding there, since nobody tested it this way, only builded
> on woody. I think it's very important to make sure security uploads
> are not going to change the package in bad ways, and suddenly building
> with a much different libc et al, and a different gcc with different
> properties apparantly (this bug is directly caused by it) might be
> resulting in such bad changes, we simply don't know, since it hasn't
> been tested.
Yes, I understand that, and I mostly agree. Now please write a lintian
warning for PT_GNU_STACK. Mass bug filing me even before a lintian
warning exists is not polite.
Reply to: