[Ltrace-devel] Welcome + argument question.

Petr Machata pmachata at redhat.com
Tue Oct 22 23:13:03 UTC 2013


Adam Malinowski <a.malinowsk2 at partner.samsung.com> writes:

>> I also see this is broken on master, probably due to my prototype
>> library work.  Clearly we don't have tests for this.  I'll have to fix
>> this before 0.8.

A fix for this is on master now, but I've pushed in the middle of
something, so even though the test suite passes, for general use it
won't even show arguments.  The reason is that I split ltrace.conf to
libc.so.conf, libacl.so.conf and libm.so.conf, but ltrace only looks for
libc.so.6.conf, libacl.so.1.conf, etc.  I'll need to change the logic a
bit so that it does the useful thing here.

> I've also looked at the source for 0.5.3 version and differences are huge.
> There was argument type STRING which is missing in current source.
> I assume that this is due to completely new approach to prototypes.

"string" is still recognized as keyword in config files, but it's not a
type anymore--it's something we call "lens".  It's a way of looking at
an array of bytes.

> I needed a tool to trace system calls blocked by seccomp filter of given
> process.
> It turned out that neither strace nor ltrace has such functionality.
> Ltrace seemed to be simpler to modify :)
> So I've added cooperation with seccomp to ltrace and enhanced summary a
> little.
> If you are interested in such changes I will give it to you.
> I'm not sure if I did it correctly but it works :)

What are the changes that you added to ltrace--can you give a short
paragraph with a description?  Much depends on whether it can be
integrated into ltrace without hacks, and without exposing
linux-specific functionality in ltrace core.  I'll also have to go and
see what seccomp is, I admit ignorance on this subject ;)

Thanks,
PM



More information about the Ltrace-devel mailing list