"Adam D. Barratt" <[EMAIL PROTECTED]> writes: > On Mon, 2008-12-01 at 13:16 -0800, Russ Allbery wrote: >> "Adam D. Barratt" <[EMAIL PROTECTED]> writes:
>>> My testing suggests that lintian may be filling up the filename buffer >>> faster than file / md5sum can empty it. Adding a 1ms sleep (well, a >>> select()), at the end of each iteration of the main processing loop in >>> the collection scripts appears to resolve the issue[1], at the expense >>> of taking over 30 seconds to check the package on a 1.8GHz amd64. >> Why would lintian not just block when writing to md5sum (or rather to >> xargs, which is what lintian is talking to directly)? xargs is >> normally used with find, and find writes as fast as it can without any >> delay. > A fair point; I did leap to a conclusion. >> It seems like something stranger is going on here. > Indeed, it would appear the delay is simply treating a symptom. The possibly related thing that changed recently is that we switched to IPC::Run from rolling equivalent code ourselves. Maybe we're breaking some IPC::Run assumption or have uncovered a latent bug? (The latter seems unlikely, since IPC::Run is fairly mainstream, but I suppose it's possible.) -- Russ Allbery ([EMAIL PROTECTED]) <http://www.eyrie.org/~eagle/> -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
