On 10/11/15 16:37, Assaf Gordon wrote: > Hello, > > (continuing old topic of csplit/errno reporting...) > >> On 23/10/15 00:40, Pádraig Brady wrote: >>> On 22/10/15 21:01, Jim Meyering wrote: >>> >>>>> I'll be able to troubleshoot and provide more information in couple of >>>>> days. >>>>> A more elaborate solution is to save the errno after 'fwrite()' in a >>>>> variable, >>>>> and print that variable. I can send such a patch if that's a better idea. >>>> >>>> Yes, this sounds better. >>> >>> Yes the tradeoff discussed above is a bad one in csplit's case >>> as it only has a single fwrite() call. > ... >> Actually it's a bit awkward to test. You might have to resort >> to tests using require_gcc_shared_ and wrap write() to return ENOSPC. > > Attached is such change with relevant test using LD_PRELOAD. > Tested on Debian with GCC 4.7.2, and on Mac OS X (where the test is skipped). > I'll soon test on more systems.
It would be good to verify we get some detail in the diagnostic, as that's what the patch is doing. I.E. something like: sed 's/:...*/.../ err compare exp err || fail=1 Otherwise looks perfect. thanks! Pádraig.
