It should be possible to at least configure SSP to always log to syslog(). Of
course since it's (apparently) /usr/lib/libssp.so.0 (usr? I should talk to my
distro maintainers about that...), I suppose I could replace this with my own
version..
The logging of SSP kills to syslog() is useful because it allows users who are
not monitoring the program in a console (think Firefox, gnome-session, etc) to
retrieve stack smash information after a crash. Currently this would allow a
user to immediately know if the program died due to a stack smash; however, in
tandem with bug #28328, this could become much more useful.
A useful consideration is that a distribution could supply a daemon that may be
configured to watch syslog() for stack smashes. This daemon could then report
stack smashes back to the distribution maintainers automatically (users rarely
do).
In tandem with bug #28328, such a tool could report not only the program that
overflowed; but also report the source file and function that was overflowed.
This is quite enough data to place a developer straight at the root of the
problem: the function that the overflowed buffer was allocated in. From there
it's relatively easy to trace down the problem-- orders of magnitude easier
than having to examine the ENTIRE codebase to figure out WHICH buffer was
overflowed AND why.
Lacking such a tool, being able to quickly determine if there was a stack smash
(and where if we consider #28328) would allow bug-reporters and newbies crying
in the IRC channel of any distribution or project to return useful information
in a few seconds, without the developers jumping through hoops trying to calm
them down. (considering bug #28328, they could be waved away in under a
minute, since they could give all useful information in that time).
--
Summary: SSP always log to syslog()
Product: gcc
Version: 4.1.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: other
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: nigelenki at comcast dot net
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28334
