Re: [PATCH] PR 9677 gcc-3.4.1-20040715-crazypaths.patch

[Andrew Haley <aph at redhat dot com>]

Ranjit Mathew writes:
 > On Wed, 18 Aug 2004 10:30:19 -0400, Bryce McKinlay <mckinlay@redhat.com> wrote:
 > > I've seen those FAILs appear from time to time too. They seem to be a
 > > bit spurious - rebuilding from a clean tree often makes them dissappear.
 > 
 > I always build GCC from a clean tree and I have been
 > seeing these failures consistently in the last one week
 > at the least.
 > 
 > 
 > > I was unable to reproduce them outside of jacks. Does anyone know of a
 > > way to get jacks to be more verbose, so that it shows the exact compiler
 > > command it uses and compiler output that it gets?
 > 
 > The output is stored in the same folder as the tests. For
 > example, for the new failure "non-jls-argument-expansion-4" you
 > will find the output (as well as the inputs) in the folder:
 > 
 >   $JACKS/tests/non-jls/argument-expansion/
 > 
 > In my case, the input comes from the file "list4" which
 > contains "C1.java\r\nC2.java" ("\r\n" is the normal CR-LF
 > pair) and the output produced by GCJ is in the
 > file "@list4.gcj.err":
 > 
 >   C1.java:0: fatal error: can't open
 >   C2.java: No such file or directory
 >   compilation terminated.

Well that one looks like DOS line ending bogosity.  We open the file
"list4" with the argument "r" (rather than "rb") so there shouldn't be
any problem with endings.  Hmm.

Andrew.

 > (These files *are* there and simply contain "class C1 {}" and
 > "class C2 {}" respecively.)

Yes, but it looks to me like the compiler is trying to open a file
called "\nC2.java".

Andrew.