[m-rev.] diff: workaround to fix nightly builds

Mark Brown mark at cs.mu.OZ.AU
Tue Aug 16 00:32:49 AEST 2005


On 16-Aug-2005, Julien Fischer <juliensf at cs.mu.OZ.AU> wrote:
> 
> On Mon, 15 Aug 2005, Mark Brown wrote:
> 
> > On 15-Aug-2005, Julien Fischer <juliensf at cs.mu.OZ.AU> wrote:
> > >
> > > Estimated hours taken: 0.1
> > > Branches: main
> > >
> > > library/Mercury.options:
> > > 	Workaround a bug in the compiler that is causing the
> > > 	library to fail to build in grade asm_fast.gc.tr.debug
> > > 	at -O5.  This is just to get the nightly builds up and
> > > 	running again.
> > >
> >
> > The following diff appears to fix the problem.  I'll commit it after
> > further testing, assuming it passes.
> >
> > Cheers,
> > Mark.
> >
> > Estimated hours taken: 1
> > Branches: main
> >
> > Fix a problem exposed by the exists_cast transformation.  saved_vars.m was
> > renaming prog_vars in procedure bodies without updating the rtti_varmaps.
> > As a result, pre-birth sets were being calculated wrongly in some cases,
> > leading to a failure of the code generator when compiliing
> > library/version_store.m at -O5 in debug grades.
> >
> > compiler/saved_vars.m:
> > 	Build up a variable renaming in the slot_info.  Apply it to the
> > 	rtti_varmaps at the end of processing.
> >
> > library/Mercury.options:
> > 	Remove the workaround for this bug.
> >
> > library/map.m:
> > 	Add a new type, renaming/1, which is a map whose key and value types
> > 	are the same.  Move two predicates for manipulating renamings from
> > 	simplify.m into map.m.
> >
> > compiler/simplify.m:
> > 	Use the map library instead of local definitions.
> >
> Could you please add the test case for this one to the test suite?
> If you arrange so that it always compiles at -O5 in a .debug grade
> at least the nightly tests on the machines that don't compile at that
> optimization level will pick it up -- in fact, compiling with
> --optimize-saved-vars-const in a debug grade should be sufficient.

No worries.  The fix failed in other cases, though, so it's back to the
drawing board.  :-(

Cheers,
Mark.

--------------------------------------------------------------------------
mercury-reviews mailing list
post:  mercury-reviews at cs.mu.oz.au
administrative address: owner-mercury-reviews at cs.mu.oz.au
unsubscribe: Address: mercury-reviews-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-reviews-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------



More information about the reviews mailing list