Hi, I'm trying to convert a PRCS repository into git using a custom fast-import frontend. It's a very large repository, and I can't convert it fast enough to stop users from committing to it while I convert it (it takes 4+ days, PRCS is old and not that quick, and I've got over 7000 commits to migrate. I've been converting from a backup). I've found that if I do an export up to a certain point, then attempt to resume the export using the marks file generated by fast-import, it complains saying (for example); error: unable to find 185643da420354e861743f3056614f4dc78c0246 fatal: object not found: 185643da420354e861743f3056614f4dc78c0246 I tried this with version 1.5.6.5-3 (+lenny2~bpo40+1, on Debian Lenny), and the offending hash and mark were in the marks file, quite a way earlier than the point from which I attempted to resume the export. I installed 1.7.1 from source and tried again and hit the same problem, although this time it seems to have trampled the marks file in the process - I had 47 entries instead of 10x that number before trying to resume. The only other information in the crash report is, "active_branches = 0 cur, 5 max". Not much to go on. I'm wondering if there are conditions where fast-import will somehow ignore a revision, or squash it into another revision, such that it accepts and processes the command stream that I've generated but doesn't generate a commit, despite it outputting a hash to the marks file? Thanks and regards, -- Rory Hunter -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html