[Bug breakpoints/24541] Incorrect evaluation of systemtap probes due to register being signed and probe expression assuming unsigned

[fche at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=24541

--- Comment #16 from Frank Ch. Eigler <fche at redhat dot com> ---

> If I am understanding correctly, the proposal here is to make both
> modifications: use %k[] on sdt.h, and also make the decoder interpret when
> the informed size is actually larger than the actual register size.

I wish we didn't have to deal with the latter, but we may not have a choice
on the latter, for larger-than-any-register types.

> I am more than fine with using %k[], of course, if it means that we will
> have the right pair of size-string@register there.

Just to confirm, there is no sign/size problem in the sdt.h machinery?

> As for the second part, I think I can accomodate and expand the current
> parser to take that into account, even though my first impression is that
> the parser shouldn't need to have this knowledge built into it.

Indeed.  And there are multiple projects that consume sdt.h macro notes
(gdb, stap, perf, several other tracing tools), so this code would to be
replicated.  We can start assembling a table in one of the wikis, or the
header file itself.

-- 
You are receiving this mail because:
You are on the CC list for the bug.