This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: dlmopen and core dumps


Some notes from the trenches.

On Wed, Dec 19, 2012 at 10:05 AM, Jan Kratochvil
<jan.kratochvil@redhat.com> wrote:
> On Sat, 15 Dec 2012 01:20:03 +0100, Carlos O'Donell wrote:
>> Why not just run a generic program, or rather a python script?
>
> Python is needlessly expensive.

In addition, we've been using embedded Python scripts to support debugging
here, and have found keeping code and python scripts in sync to be a bit
of a challenge.

Writing an automated GDB test that verifies the "sync" state is possible;
requiring a working GDB every time you perform a "make check" in glibc is
not really plausible, especially when bootstrapping on a new platform.

> There does not exist only GDB, there exist
> other smaller tools examining external programs which can for example run on
> embedded targets.

And don't forget self-inspecting programs. When running on a cluster, it is
really handy for SIGSEGV and SIGABRT handler to be able to log symbolized
stack trace, which requires access to the list of loaded libraries [1].

> Why not to use DWARF expression code?

Sounds like an excellent idea to me.

[1] The list can be obtained by parsing /proc/self/maps, except /proc may
not be mounted, or may not be present in the chroot jail, or may be made
inaccessible by sandboxing policy.

-- 
Paul Pluzhnikov


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]