Mike Frysinger wrote:
> On Wednesday 10 December 2008 20:34:03 Jiří Paleček wrote:
>> the tst_brk function doesn't print anything when Tst_count>=TST_TOTAL.
>
> tests shouldnt be calling tst_brk() when tst_count == TST_TOTAL, and tst_count
> should never get above TST_TOTAL. we should add a sanity/assert test to
> tst_exit() for this i think.
>
> what tests are you seeing this with ?
It was the utimensat01, which declare TST_TOTAL=0, which caused breakage when I
changed a tst_res to tst_brk. However, other tests are not much better - eg.
signalfd declares TST_STOTAL=1 although I see two tests there, and IIUC,
tst_brk
in the cleanup function are (almost) always called with Tst_count==TST_TOTAL.
I think if tst_res ignores wrong Tst_count, tst_brk should too, because it is
easier than auditing all code paths for this kind of correctness, easier by far.
>> See the patch for comment.
>
> you're clearly using `git` ... why arent you using `git send-email` ? that'll
> post the patch properly inline so it can easily be reviewed. sticking a patch
> in an e-mail and then attaching that e-mail to another e-mail isnt easy to
> review ...
There are multiple reasons, like "git send-email" doesn't work for me, it
avoids
whitespace/line breaking troubles when read from some web archives, prevents
excessively long replies quoting the patch text, alows me to stick patch to a
reply, etc. But if it would help, I could eg. publish the repository and send
references to commits... or something
Regards
Jiri Palecek
------------------------------------------------------------------------------
_______________________________________________
Ltp-list mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ltp-list
