-fcheck-memory-usage, Checker, Purify, and core files.

-fcheck-memory-usage, Checker, Purify, and core files.

[Mike Simons]

  Purify with EGCS on our Sparcs is making a royal racket, but 
successfully links everything I've wasted time one when given the 
correct options, (I hope to send tips explaining how to get it working 
or at least what works for me).

So I looked into the new -fcheck-memory-usage option...

  When even a simple program (one malloc) is compiled with this option it 
causes an internal compiler error 11 (SEGV), and dumps a 10 Meg core
file (this has been reported by at least two people already).  There 
is no text documentation on this option, so I grep'ed and scanned the 
EGCS source for info, not knowing anything about the compiler internals
there wasn't much gleaned.  From the change log I see patch that
added this option was submitted into the GNU baseline August 2nd 1997 
by Tristan Gingold, (the author of Checker).
  On August 3rd EGCS branched off.  I suspect that more changes where made 
to the GNU baseline related to this option since then, but it hasn't 
been copied over.  (At that time I could only speculate... now that
GGCS 2.8.0 is out I can check for myself).

  Tristan said that all of his changes were submitted to the GNU group 
and accepted, he hasn't submitted to EGCS.  Also that this option is 
relies on the next version of Checker, which should come out soon.

  Could someone with access to the GGCS repository merge any additional
patches into EGCS?  I will try looking at 2.8.0 if no one speaks up... 
yeah learning curve.
  Meanwhile, could the command line parser engine for EGCS simply say that
this option is unsupported at this time (give a date)?  So others don't 
waste time figuring out why the compiler cores...

-- 

    Thanks,
      Mike Simons
      Science Applications International Corporation
      msimons@saic1.com                  703-925-5674

Re: -fcheck-memory-usage, Checker, Purify, and core files.

[Jeffrey A Law]

  In message < 199801162341.SAA09526@aura.saic1.com >you write:
  >   When even a simple program (one malloc) is compiled with this option it 
  > causes an internal compiler error 11 (SEGV), and dumps a 10 Meg core
  > file (this has been reported by at least two people already).
I don't think -fcheck-memory-usage has ever worked, in any tree.

  >   On August 3rd EGCS branched off.  I suspect that more changes where made 
  > to the GNU baseline related to this option since then, but it hasn't 
  > been copied over.  (At that time I could only speculate... now that
  > GGCS 2.8.0 is out I can check for myself).
We do periodic merges from the gcc2 tree, which includes the -fcheck-memory-usage
code.  The latest snapshot has everything you'd find in gcc-2.8.

However, I'd still be suprised if -fcheck-memory-usage works.

jeff

Re: -fcheck-memory-usage, Checker, Purify, and core files.

[Martin von Loewis]

>   Purify with EGCS on our Sparcs is making a royal racket, but 
> successfully links everything I've wasted time one when given the 
> correct options, (I hope to send tips explaining how to get it working 
> or at least what works for me).

As for Purify and egcs, I found that the least painful approach is to
update Purify. They've released 4.1 for Solaris recently, and it does
not complain about the many N_FUN symbols, and it also does not crash
so often, anymore.

Just check out the Rational home page, and navigate tot the Purify
download section. Have your version number and password ready.

Hope this helps,
Martin