On Fri, Dec 12, 2014 at 7:24 PM, KO Myung-Hun <[email protected]> wrote: > > > Jim Meyering wrote: >> On Wed, Dec 10, 2014 at 9:37 PM, KO Myung-Hun <[email protected]> wrote: >>> Jim Meyering wrote: >>>> On Tue, Nov 25, 2014 at 6:29 PM, KO Myung-Hun <[email protected]> wrote: >>>>> Hi/2. >>>>> >>>>> These are OS/2 patches. >>>>> >>>>> Review, please... >>>>> >>>>> [PATCH 1/2] build: use quotation mark(") for PATH >>>>> [PATCH 2/2] diff: skip test if seek is not possible on OS/2 kLIBC >>>> >>>> I saw no patches here, but was able to dig them out of the >>>> bug-tracking system at http://debbugs.gnu.org/19186. >>>> I have applied and pushed the first one. Thank you. >>>> >>>> However, for the second, my inclination is that if at all possible, >>>> it should be fixed via an lseek replacement that does something >>>> more sensible. Of course, that may not be possible, but from >>>> what you've presented so far, I cannot tell. Please demonstrate >>>> a use of diff that shows how OS/2+kLIBC's lseek fails. >>> >>> Do you want this ? >>> >>> ----- >>> $ cat file | diff - file >>> diff.exe: -: Invalid seek >>> ----- >> >> Yes. Thanks. That shows how lseek-pipe fails on your system. >> Please adjust your patch to make diff ignore failure only in that case, >> i.e., when errno == EINVAL (assuming strerror(EINVAL) >> produces that diagnostic). > > Do you mean to add errno == EINVAL to if() for pfatal_with_name() ?
No. Skip the test only when this lseek fails with precisely that errno value: +#ifdef __KLIBC__ + /* Skip test if seek is not possible */ + skip_test = skip_test || lseek (current->desc, 0, SEEK_CUR) < 0; +#endif
