[patch, gfortran] pr22570 and related issues.

Paul Thomas paulthomas2@wanadoo.fr
Tue Jul 26 17:49:00 GMT 2005


The enclosed patch, testcase and changelog entries, hopefully deal with 
the last of the issues arising from the fixes to the NIST triggered bugs 
in formatted io.

Thanks to Jack Howarth, 
http://gcc.gnu.org/ml/fortran/2005-07/msg00395.html, for pointing out 
that there were still trailing spaces getting through, in spite of all 
the counter measures.  Most of this patch fixes that problem.  At the 
same time, an incorrect treatment of slash formatting had broken fixed 
record length io and the patch fixes that too.

One thing that is apparent from the last couple of weeks is that the 
testsuite tends only to test that which was previously broken.  Often 
this is sufficient.  However, with formatted io, it turned out to be too 
easy to break that which was previously unbroken; not in small part 
becuse there are so many possible variations of what can be done with 
formatted io.  For this reason, the enclosed test case is rather more 
exhaustive than it need be.   Something similar is needed for fixed 
record length io.  In the longer run, there might be some advantage in 
modifying the NIST suite to run as part of the gfortran testsuite.  The 
work required to do this would be perspiration rather than inspiration 
based.   If it is thought to be useful, we could look into this, 
although there are no promises on the timescale.

The patch is regtested (including enclosed testcase) on RH9/Athlon 1700. 
 It also seems to fix all the problems except that of array internal 
files in the NIST suite.

OK to submit the patch to mainline and 4.0?

For the testcase, we invite comments for improvement.  If it is felt to 
be useful to commit it right away, this can be done.

Regards

Jerry DeLisle and Paul Thomas


-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: transfer_correction.msg
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20050726/1a3451f2/attachment.ksh>


More information about the Gcc-patches mailing list