[Pkg-clamav-devel] Processing of clamav_0.98.4+dfsg-1_i386.changes

Scott Kitterman debian at kitterman.com
Sat Jun 21 16:33:53 UTC 2014


On Saturday, June 21, 2014 17:59:43 Andreas Cadhalpun wrote:
> Hi,
> 
> On 19.06.2014 00:18, Scott Kitterman wrote:
> > On June 18, 2014 3:50:50 PM EDT, Andreas Cadhalpun 
<andreas.cadhalpun at googlemail.com> wrote:
> >> Did you intentionally re-include the libclamav/c++/llvm directory?
> > 
> > No.  Forgot.   It might be nice to have a get-orig-source rule if someone
> > has time.
> We have the split-tarball.sh, which repacks the tarball correctly.
> 
> It wouldn't be difficult to add a get-orig-source rule that calls uscan
> and then split-tarball.sh, but something is currently broken, because
> uscan doesn't find the 0.98.4 tarball.
> Indeed the page [1] where uscan looks doesn't contain a link to this
> tarball.
> 
> By the way, I added two patches to fix building on Hurd:
> 0009-Fix-compiling-on-Hurd.patch:
> This should be upstreamed, but I didn't do that yet, because it doesn't
> work alone.
> 
> 0010-Workaround-a-bug-in-libc-on-Hurd.patch:
> This should be dropped, once the libc bug [2] is fixed.
> Then the other patch can go upstream.

I had a good discussion yesterday with a couple of the upstream developers.  
Here's what we came up with:


0001-Change-paths-in-sample-conf-file-to-match-Debian.patch

I asked them if they would consider a configure option to set these values so 
we wouldn't have to do it via patch.  They agree that was a reasonable 
request, so I'll file a bug.

0002-Fix-STAT64-definition-and-add-missing-includes.patch

They didn't agree with this patch.  I checked and i386 now builds without it, 
so I dropped it.  As written, the patch may affect large file scanning since 
(according to them) on 32bit systems glibc doesn't handle switching to STAT64 
despite claims that it does (they specifically disagreed this was Windows only 
code).

0003-Add-an-additional-n-after-the-number-in-the-pidfile.patch

They agreed to upstream this.

0004-unit_tests-increment-test-timeout-from-40secs-to-5mi.patch

I asked for some kind of a switch to control this so we could do it per-
architecture if needed.  They agreed and in fact had run into some similar 
issues in their powerpc testing.  They plan to add this as a make variable 
that we can override if needed.

0005-Fix-FTBFS-with-LLVM-3.1-3.4.patch
0007-Fix-underlinking-with-llvm-in-libclamav-c-Makefile.a.patch

They have a branch that uses system LLVM that used our patches as a starting 
point.  This may or may not make 0.98.5 as they have other things they are 
working on for 0.98.5 and limited time.

0006-enable-llvm-jit-on-kfreebsd.patch

They will probably change this to be *bsd* since all BDSs can support 
JIT/LLVM.  I asked about other archs where we have LLVM and their may problem 
was lack of hardware to test.

My dropping of the STAT64 patch did collide with your changes.   I think I 
fixed it up in git, but please check after you pull again.

Thanks,

Scott K



More information about the Pkg-clamav-devel mailing list