Hi! > Thanks. > > RDFLib 3.0.0 was recently released. Thus, before uploading a patched > debian/rules, I'd like to test whether this issue is reproducible with > the new version as well. I'm preparing a package, it would be great if > you could test it on your mipsel virtual machine. Give me some time, > you will have it in your inbox shortly.
Yes, of course. > Anyway, this is obviously a workaround and, as you pointed out before, > we need to generate a minimal test case to report a bug > against gcc. The problem is that it seems impossible to really make a short test case. The parser contains many if-blocks with the condition if(self->verbose). I removed them because self->verbose is always false in our case. The bug no longer occurred. Then I did a binary search on the number of blocks I can remove before the bug disappears. I now have two versions of the parser that only differ in two lines, but one exposing the problem, the other not. My guess is that there is some wrong overflow detection in gcc when relative addressing is used (that would explain why it only occurs on long functions). However, I need to further investigate this, and even if I can find the root of the problem, it will take some time that a fix is made and included in Debian's toolchain. WM -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
