[Patch, libgfortran] PR16435 and NIST fm908.for

[Paul Thomas]

This patch fixes two patches that were sufficiently interwoven as to 
warrant fixing together.

The first is that described in PR16435 for which there is already an 
XFAIL testcase, f77-edit-x-out.f.  Here the problem was that trailing 
spaces were output if a formatted write terminated with X or T-editting.

The bug triggered by fm908.for, which is awaiting the fix for internal 
io to arrays to be 100% successful, is due to spaces being written over 
already output data if a left tab was followed by a right tab. The 
enclosed testcase illustrates what I mean.

The patches, ChangeLog entries and the testcase are attached.

Whitespace differences have been suppressed because I took the 
opportunity to clean up transfer.c a bit.

(Tobi, Before you point out the potential use of MAX, I need to find out 
why MIN does not work in the previous patch.  Do these macros have 
side-effects?  Take it as read, that if I figure out the problem, I will 
use both!)

Regtested on RH9/Athlon

                === gfortran Summary ===
 
# of expected passes            8036
# of unexpected successes       8    << g77/f77-edit-x-out.f
# of expected failures          9
# of unsupported tests          8

OK for 4.1 and 4.1?

Paul T


-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: transfer.msg
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20050712/3cbf2b8b/attachment.ksh>