Bug#818148: bullet: Out of memory errors do not make the build to stop

[Markus Koschany]

Control: severity -1 wishlist
Control: tags -1 - moreinfo + unreproducible

Am 15.03.2016 um 14:41 schrieb Santiago Vila:
> On Tue, Mar 15, 2016 at 02:10:34PM +0100, Markus Koschany wrote:
>>> Those are errors that should be caught and the build should have
>>> failed, but they are not.
>>
>> Why should we make the build fail, if the build log shows that Bullet
>> has been built successfully?
> 
> Because everything seems to indicate that this is a bug in the Makefile.
> 
> An example:
> 
> sometarget:
> 	command1; command2; echo successful
> 
> The fact that "make sometarget" says "echo successful" does not mean
> every command was executed successfully.
> 
> It could be that command1 fails and command2 hides the error. So it
> says "sucessfull" improperly.
> 
> In this example, the right Makefile should be like this:
> 
> sometarget:
> 	command1 && command2 && echo successful
> 
>> Is there anything different with your
>> compiled version of Bullet and the current one in Debian's archive?
> 
> As I'm only interested in checking that "dpkg-buildpackage -A" works,
> I throw the binaries after the build, so I don't know.
> 
> But it is easy to imagine that a process that says "out of memory" is
> unlikely to produce the correct result.
> 
> If a process says "out of memory" and the output is the same,
> it means whatever does the process is not used at all.
> 
> If you don't mind about the result, why run a command at all?
> 
> As I said before, everything seems to indicate that there is an
> underlying Makefile bug here. I didn't investigate any further because
> there is plenty of evidence ("out of memory" + "successful" build),
> that's why I submitted this bug.
> 
> Thanks.

Thank you for providing more information. I would still recommend that
you rerun your tests without the parallel option in case you are running
out of memory and try to investigate if your issue still persists.

I am going to downgrade the severity to wishlist because I think this is
an improvement suggestion but not an important bug for Debian as long as
Bullet builds fine on Debian's build servers and also on my own system.
Apparently this is also more related to graphviz' dot program and
doxygen in general. As I mentioned before there are available CMAKE
options to adjust the build process for your specific host system and
you could also simply increase your swap space. I consider this issue to
be very low priority for myself but of course I won't stop anyone else
from working on this.

Regards,

Markus


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 949 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-games-devel/attachments/20160315/fc2d92d7/attachment.sig>