Re: qemu version used on sh4 buildds (was: Bug#914323: file: non-zero exit code for shared objects on sh4)
Hello!
On 11/24/18 9:08 AM, Christoph Biedl wrote:
> Long story short: The file(1) program caused the build errors on sh4 for
> several packages. After *many* hours of digging I found the pread64
> system call used by libmagic isn't properly translated when using the
> stretch version of qemu-user-static (1:2.8+dfsg-6+deb9u5). The
> buster/sid version has upstream's fix¹, and I can confirm file(1) will
> work as exepected then. So, not a bug in the file package, closing.
>
> The thing that puzzles me: James Clarke (jrtc27, cc:'ed) stated the
> buildds use a manually built and fairly recent qemu version. These two
> bits of information don't match.
I am in charge of this, not James. What most likely happened is that an
apt upgrade in stable also updated qemu (security fixes) and then a
version without the patch silently slipped onto the buildds.
Since I was very busy the past weeks with SUSE work, I wasn't able to
take care of the buildds and hence I didn't notice the issue. I have
updated qemu to a very recent version and the issue has been resolved,
I also had to manually build "apt" because the first 1.8.0 version had
a serious regression which made it lock up this resulted in buildds
to hang.
> As I neither have access to the buildds nor am interested in getting
> one: Can you please check the problem is really fixed there? Should
> be as simple as running "file /bin/ls" in a qemulated sh4 chroot.
There are no qemulated chroots, we are using the tarball setup from
buildd.
> Expected result: /bin/ls: ELF 32-bit LSB executable, Renesas SH, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux.so.2, BuildID[sha1]=bc61964767543a3ac5b756454ac96634c24a9f1b, for GNU/Linux 3.2.0, stripped
>
> Broken: /bin/ls: ERROR: error reading (Invalid argument)
> (also, non-zero exit)
I'm aware of the implication of this issue. This is how the qemu bug
was discovered in the first place.
In the future, please contact James or me directly, either through mail
or on #debian-ports in OFTC IRC as in most cases we already know the
problem.
Thanks for taking care in Debian Ports!
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@debian.org
`. `' Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
Reply to: