Patch: DIR_SEPARATOR usage in jcf-write.c and jcf-io.c

[Andrew Haley <aph at redhat dot com>]

Ranjit Mathew writes:
 > Hi,
 > 
 >     This is a resubmission of the DIR_SEPARATOR patch for jcf-io.c 
 > and jcf-write.c in the front-end.
 > 
 > The last time I had submitted this, it was suggested that we 
 > might be able to use something from libiberty instead - AFAICT, 
 > the only relevant functions in libiberty seem to be the 
 > basename( )/lbasename( ) functions. AFAIK, if we are to use them, 
 > the code will still be as "untidy" as this, as they do not 
 > directly address the issue we are trying to resolve here.
 > 
 > This patch resolves the bug on Win32 where "gcj -C foo\Bar.java"
 > produces the class file in the current directory instead of in
 > "foo" while "gcj -C foo/Bar.java" has the correct behaviour.
 > 
 > Ranjit.
 > 
 > 
 > Index: ChangeLog
 > from  Ranjit Mathew  <rmathew at hotmail dot com>
 > 
 > 	* jcf-io.c (caching_stat): Account for both DIR_SEPARATOR
 > 	and DIR_SEPARATOR_2 for a target.
 > 	Correct minor typos.
 > 
 > 	* jcf-write.c (make_class_file_name): Take both DIR_SEPARATOR
 > 	and DIR_SEPARATOR_2 for a target into account.

This is a compiler patch, so it should have been sent to gcc-patches
as well as java-patches.

Also, I had to fix a warning.

Applied to mainline and 3.3 branch.

Andrew.