Opened 13 years ago

Closed 9 years ago

#40 closed task (fixed)

Is --fomit-frame-pointer worth it?

Reported by: lou Owned by: andreasw
Priority: minor Component: configuration tests
Version: 0.5 Keywords: --fomit-frame-pointer
Cc:

Description

The occasional bug report trickles in to the effect that C++ exception handling fails for some version of GCC (the 3.x series seems vulnerable). Usually it's reported as a failure in one of the unit tests. When pursued, the problem turns out to be this: The optimised build specifies --fomit-frame-pointer, and this tickles some bug in GCC which causes a failure to catch a throw.

A workaround is posted to the BuildTools current issues page. On the other hand, we'll likely keep on seeing this trickle of bug reports. Is the gain from --fomit-frame-pointer worth the low level irritation?

Change History (3)

comment:1 Changed 13 years ago by andreasw

I included that flag because John Forrest used it and said that you get some speedup from it (not huge though). Can the error easily be tested? Is there a simple example program that could be compiled and run during configuration?

comment:2 Changed 12 years ago by andreasw

Stefan Vigerske just told me that in Ipopt, using GCC 3.5.5, he uses -fno-omit-frame-pointer -momit-leaf-frame-pointer, and with that he was able to run. We could switch to that...?

Just a thought

comment:3 Changed 9 years ago by stefan

  • Resolution set to fixed
  • Status changed from new to closed

Since -fomit-frame-pointer still seem to make trouble on cygwin/gcc, and since the gcc man page on Linux claims that it is automatically set in optimized mode, this flag has been removed from the compiler flags with chgset 1581 (BuildTools?/trunk).

Note: See TracTickets for help on using tickets.