BFS corrupted during very large project build

[haiqu]

I have a repeatable fault when at the linking stage of a very large project. The GNU linker LD reports that the filesystem has become R/O at the final linking stage and I have to run chkbfs on the drive from Dano to correct the problem.

[axeld]

Can you provide at least a bit more info? Like what project it is, and maybe a link to it in case it's freely available. A reproducible file system bug would really be nice!

[haiqu]

Tried to add data to this report yesterday but the ticket was locked somehow, sorry.

The project was in Pascal so the tools aren't readily available for you to test a build. What I *can* add is that the crash hit EXACTLY as the file size passed a size of 64Mb, which is starting to look like a boundary condition problem. I might also add that my earlier report #3808 about problems creating and copying files and/or writing to CD-ROM occurs at around 32Mb.

[haiqu]

While I'm here, I should add that chkbfs run from Dano reported that there were NULL inodes and it took a bit of work to fix the drive. I had to delete large chunks of my work and run chkbfs three times to be sure it was all good again. I was ready to init the partition and start again at one point.

Re-running the compilation on a faster machine caused NO p[roblems, and the eventual executable file size was a whopping 84Mb, is that a record?

[axeld]

As this info is not really helpful, I'll just regard it as a duplicate of #3150.