Re: [Libunwind-devel] cursor.is_interrupted() or is_...

[Arun Sharma]

On Jan 29, 2008 2:36 AM, Mark Wielaard <address@hidden> wrote:

Also related to this discussion are the following set of patches in the

frysk libunwind tree to better support signal frames. They don't expose
a new interface though, they just handle the above issue better. Do read
the following two bug reports which contain some more discussion on the
subject (and show that older versions of glibc didn't provide full
unwind info in all cases, Jan also pushed a fix for that upstream).

    RESOLVES=http://sourceware.org/bugzilla/show_bug.cgi?id=3529
    RELATED=https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=217087

I like the patch in general. Hopefully it'll make the as->validate = 1 and the associated msync() hack go away.

Some comments:

* Could we do some configure time checks for the UNW_LOCAL_ONLY case to see if the glibc is patched. If it is, perhaps we can simplify unw_is_signal_frame?

* unw_step has some lines which exceed 80 columns

* There is a new file with HP's 2006 copyright?

* The patch seems to undo:

http://git.kernel.org/gitweb.cgi?p=libs/libunwind/libunwind.git;a=commit;h=31440e9796bb34146372df52ed59c4f68ea5839d

why?

* I'll let David comment on the new API being introduced: tdep_fetch_proc_info_post()

 -Arun