introducing a stable branch
robert at ocallahan.org
Tue Apr 21 22:11:37 UTC 2015
On Wed, Apr 22, 2015 at 6:10 AM, Steve Fink <sphink at gmail.com> wrote:
> reverse-exec is the backend for gdb/rr's reverse-continue and friends? I'm
> on a somewhat old checkout. It failed for me the first time I tried, which
> I couldn't reproduce, and it's worked perfectly the dozen or so times I've
> used it since then.
> And it's *awesome*. I had two intermittent failures, and settled into the
> procedure of:
> % while rr my-command; do echo passed; done
> % emacs
> M-x gud-gdb, set command to rr replay
> gdb> set annot 1
> gdb> run
> feel happy
> But these were JS engine only, which may be why I didn't hit any bugs. I
> did have a third bug that would not reproduce under rr. It was related to
> thread scheduling. It made me wonder if rr while recording could force
> thread execution to be interleaved in obnoxious ways. (Or even have a
> client directive that you could compile in to request a context switch at a
> particular point.)
You can use "rr record -cNNNN" to get more frequent context switches.
(Default value of NNNN is 250,000.) You can add "sched_yield()" calls to
your code to force a context switch.
Please collect examples of bugs that you've solved that rr couldn't
reproduce, and file rr issues on them. Then we can study how to fix rr to
oIo otoeololo oyooouo otohoaoto oaonoyooonoeo owohooo oioso oaonogoroyo
owoiotoho oao oboroootohoeoro oooro osoiosotoeoro owoiololo oboeo
osouobojoeocoto otooo ojouodogomoeonoto.o oAogoaoiono,o oaonoyooonoeo
osoaoyoso otooo oao oboroootohoeoro oooro osoiosotoeoro,o o‘oRoaocoao,o’o
oaonosowoeoroaoboloeo otooo otohoeo ocooouoroto.o oAonodo oaonoyooonoeo
osoaoyoso,o o‘oYooouo ofooooolo!o’o owoiololo oboeo oiono odoaonogoeoro
otohoeo ofoioroeo ooofo ohoeololo.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the rr-dev