Re: Shuffler revisited
Stephen R. van den Berg wrote: >With regard to the Shuffler: the Pike 8.0 shuffler is very fragile >and could (at any time) result in datacorruption and/or segmentation faults. >Even though, considering the incremental discovery of all the >errors in the Shuffler feels a bit like whack-a-mole, I can reason that >most likely the grave memory corruption bugs are gone, considering >that all callback connections between Shuffle and sources have now >been covered and rewritten. I checked the 8.0 version of the Shuffler. There are some differences, not sure if they present a problem backporting. The simplest would probably be to replace the whole 8.0 version of the module with the 8.1 version; the question is if that would be a viable option (or even works without too much trouble). I personally have no need for a working Shuffler in 8.0, so I'll leave it untouched. -- Stephen.
Re: Shuffler revisited
Tobias S. Josefowitz @ Pike developers forum wrote: >>:-). Well, I don't think these problems touch the Shuffler. >Oh I see, I thought FakePike would be using Shuffler... No, indeed. I fixed the FakePipe issue. It passes all tests, AFAICS. With regard to the Shuffler: the Pike 8.0 shuffler is very fragile and could (at any time) result in datacorruption and/or segmentation faults. Even though, considering the incremental discovery of all the errors in the Shuffler feels a bit like whack-a-mole, I can reason that most likely the grave memory corruption bugs are gone, considering that all callback connections between Shuffle and sources have now been covered and rewritten. -- Stephen.
Re: Shuffler revisited
>:-). Well, I don't think these problems touch the Shuffler. Oh I see, I thought FakePike would be using Shuffler...
Re: Shuffler revisited
Tobias S. Josefowitz @ Pike developers forum wrote: >I am still seeing quite a few testsuite failures, like: >Doing tests in modules/_Stdio/testsuite (181 tests, pid 12880) >Socket test >Child: Copying 19712 bytes of data on 14 fake pipes Failed to read >complete data, errno=0, "Success". >0:Input buffer: >0:Expected data: 1536 bytes. Those are probably related to me or someone else rebreaking the FakePipe. I'll check it. >Call me old-school, but I would not necessarily think of backporting >to the stable branch while that is going on... :-). Well, I don't think these problems touch the Shuffler. -- Stephen.