[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: clamav failure on hurd-i386


On 06.07.2014 13:37, Samuel Thibault wrote:
Andreas Cadhalpun, le Sat 05 Jul 2014 19:54:03 +0200, a écrit :
clamav 0.98.4+dfsg-2 compiled successfully on hurd-i386, but then the build
failed due to test-suite errors [1].
The reason for these is that the permission to create a local socket to test
the communication between clamd and clamdscan was denied. This works fine
for other architectures and also when testing in a hurd virtual machine, so
I assume there is a problem with the buildd.

Mmm, actually it is most probably a problem with chrooted hurd builds.
In chroots local sockets don't work properly, one currently has to
use the unchrooted path.

That's likely the problem.

We usually work around it by making /tmp
bind-mounted into the chroot, so that the chrooted path works unchrooted
too. A workaround for clamd would be to put the local socket somewhere
in /tmp instead of inside the build tree.

I'm afraid that can't be done easily, as it would break parallel builds.
The problem is that there are three tests using the path for this socket. The tests are run in different directories, so using a local path for the sockets is mandatory or the tests would interfere on parallel builds.

The buildd could also have its
unchrooted /build bind-mounted inside chroots as /build, so that the
local socket path would work.

Given the above this seems like the simpler solution.

Best regards,

Reply to: