apex/tjck*.sac crash in CSE

Bugzilla Link	955
Created on	May 15, 2012 20:40
Resolution	FIXED
Resolved on	Jul 07, 2012 22:26
Version	svn
OS	Linux
Architecture	PC
Attachments	tjck.sac, crud.breaks.sac, crud.sac

Extended Description

Created an attachment (id=887)
source code to reproduce fault
These benchmarks have been broken for some time now, so recent work
on LUR/LIR is probably not the problem, as shown by the fact that
the crash happens with these arguments:
sac2c crud.sac  -v3 -O3 -v4 -d treecheck -chkfreq 4   -doawlf -maxwlur 1  -nodlir  -nowlir -nolur -nowlur
No crash unless -doawlf is enabled...
sac2c -V
sac2c v1.00-beta (Haggis And Apple)
 developer rev 17818:MODIFIED linux-gnu_x86_64
 (Tue May 15 09:02:34 EDT 2012 by sac)
The treecheck shows several potential troublemakers:
           -> Running syntax tree checks
           Inserting index vector extrema ...
           -> Running syntax tree checks
WARNING: AVIS_SSAASSIGN hasnt the right type. It should be: N_assign
WARNING: AVIS_SSAASSIGN hasnt the right type. It should be: N_assign
           Inserting symbolic array attributes ...
           -> Running syntax tree checks
WARNING: AVIS_DECL hasnt the right type. It should be: N_vardec
          Inferring algebraically foldable with-loops ...
           -> Running syntax tree checks
WARNING: WITH_REFERENCED_CONSUMERWL hasnt the right type. It should be:
WARNING: N_avis
           Applying with-loop scalarization ...
           -> Running syntax tree checks
WARNING: AVIS_SSAASSIGN hasnt the right type. It should be: N_assign
WARNING: AVIS_SSAASSIGN hasnt the right type. It should be: N_assign
          -> Running syntax tree checks
           Applying withloop unrolling ...
  WLUR: -maxwlur 3 would unroll fold with-loop
           -> Running syntax tree checks
WARNING: AVIS_SSAASSIGN hasnt the right type. It should be: N_assign
WARNING: AVIS_DECL hasnt the right type. It should be: N_vardec
Some of these may be false alarms, due to incorrect specification
of legal AST contents in ast.xml. However, I am assigning this bug to 
me, and I'll take a crack at fixing up the -doawlf complaints
and see if that fixes the problem.
The immediate cause of the crash is a failure to find a name
in the PropagateIdenticalReturn2Results search.