target/3037: ARM port: Wrong flow info after reload

[rth@gcc.gnu.org]

Synopsis: ARM port: Wrong flow info after reload

State-Changed-From-To: open->feedback
State-Changed-By: rth
State-Changed-When: Mon Jun 11 14:34:36 2001
State-Changed-Why:
    Err, ok, yes, the interaction between the arm backend and flow
    keeps registers live when they oughtn't be.  That seems rather
    minor though.  Why is this pr marked "critical"?  I'm presuming
    based on "ice-on-legal-code" that this test case was supposed 
    to segfault or something, but as you didn't give compilation 
    flags I was forced to guess, and didn't come up with a
    combination that dies.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=3037&database=gcc