Hello Stephen, Am 31.05.20 um 22:52 schrieb Stephen Kitt: > On Sun, 31 May 2020 18:03:01 +0200, Bernhard Übelacker > <bernha...@mailbox.org> wrote: >> Unfortunately I found recording not working in a small test [1]. >> >> A git bisect led to a helper function introduced by upstream in [2]. >> This helper uses a function parameter of type off_t. >> But the pwrite64 from glibc uses off64_t. >> Therefore offset values with the highest bit set get incorrectly >> converted to the 64 bit type, e.g.: 0xbffff622 -> 0xffffffffbffff622 >> Therefore the write fails and rr stops. >> >> Attached patch makes rr work on i386 and passing most tests >> in my test VM, except [3]. Have not tested for side effects at amd64. >> >> I am going to try to forward this to upstream too, >> will update this bug then. > > Ah, I assumed it would work, and I didn’t even test, let alone run the test > suite. > > Thanks for the patch, I’ll see if the failing tests mean anything to me... > > Regards, > > Stephen
I reported to upstream this issue: https://github.com/mozilla/rr/issues/2597 Let's see what they think about it. The failing tests might be related to me using a VM maybe ... To do the make test I had to link rr_page_* and rr_exec_stub from rr-5.3.0/debian/rr/usr/lib/rr/ to rr-5.3.0/build/bin/../lib/rr/. Therefore I don't know if the failures might be a result of just some more files missing not found ... Kind regards, Bernhard
