On Mon, Oct 3, 2011 at 3:55 AM, Nicolas Chauvet <span dir="ltr"><<a href="mailto:kwizart@gmail.com" target="_blank">kwizart@gmail.com</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
2011/10/3 solarflow99 <<a href="mailto:solarflow99@gmail.com" target="_blank">solarflow99@gmail.com</a>>:<br>
<div><div></div><div>> I have a strange problem here, my package will build fine, but if anyone<br>
> tries to run it, the dam thing just hangs.<br>
> I can build in mock on my own system and it runs just fine. Others are<br>
> starting to report this bug against the package so know its not just me.<br>
><br>
> <a href="https://bugzilla.rpmfusion.org/show_bug.cgi?id=1959" target="_blank">https://bugzilla.rpmfusion.org/show_bug.cgi?id=1959</a><br>
><br>
><br>
> I can compare the binary is different from rpmfusion and my own build:<br>
><br>
> 1828572 Oct 3 02:39 /usr/bin/dosemu.bin<br>
> 1834428 Sep 9 01:07 /usr/bin/dosemu.bin - The bad one from rpmfusion<br>
> build<br>
><br>
> The build logs seem almost identical. Does anyone know why there might be<br>
> something different? Are all updates applied on the builder? I see some<br>
> minor version differences, however the bad binary was compiled almost a<br>
> month ago.<br>
><br>
> Any suggestions at all?<br>
<br>
</div></div>Does this appear on all target or only for F-16 ?<br>
Can you resubmit a job?<br>
build.log can be the same but root.log are probably not. That's a more<br>
probable path before to consider a problem on the builder.<br>
<br>
Right now we have the updates-testing repository enabled. This<br>
probably shouldn't be kept.<br>
This has helped a lot when building kernel modules and others packages<br>
that would have needed an override, but has leaded to issues with<br>
packages hardcode a matching version (perl qt mostly).<br></blockquote><div><br><br>Hi, thanks for your suggestions, it looks like it was rpmfusion just happened to have a difference in CFLAGS used, which happened to trigger a bug in this package. Dosemu uses some attribute((constructor)) functions that get called before main. The order isn't known (for newer GCCs you can set priorities but they are not used). Somehow the RPMFusion build ended up with an order that triggered the hang (that has, in principle been present in dosemu for many years), a linked-list where an element referenced itself.<br>
<br>Is there any way to see what URL's are being used by mock for the builder? I want to build in mock with exactly the same repos as rpmfusion is using.<br><br><br></div></div>