This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [3.3/3.4 PATCH] fix broken test g++.dg/template/spec10.C


On Monday 06 October 2003 16:23, Bob Wilson wrote:
> > The simulator (or deja, if done in there) should do status&255 on the
> > return value from main before passing back to the environment if they
> > want to match UNIX more closely.  This is ultimately why you saw the
> > problem and others don't.
>
> I had never noticed before that exit is only supposed to return the low 8
> bits of the status to the parent.  I'll fix it in our simulator.

Replying to my own message....

It turns out that the Xtensa simulator has nothing to do with this.  Deja's 
testglue.c wrapper for exit() prints out the full int value of the status.  
I'm glad you pointed out why this problem wasn't exposed on Unix targets, but 
I'm not inclined to pursue changing the testglue code since I'm not convinced 
that it's wrong.  It's different than Unix but not wrong.  The test program, 
on the other hand, is definitely wrong since all execution tests are required 
to have an exit code of 0 on success.

--Bob


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]