Under redhat 5.1, I found and fixed the following bug, which I emailed to bugs. Here is my bug description: ----------cut---------------------------------------------- Hello: I have found a bug in lpd-0.31. It seems that when running linux as a print server on a PC and you direct another machine running lpd (I have tried this with another linux PC and a DEC alpha OSF system) to send its print jobs to your print server, the jobs arrive in the queue, but are not printed. When subsequently a local user prints a job, both the local and the previously queued remote job print. I tracked this down to a buffer for file names of remote jobs which was too small. This caused the child spawned by lpd to print the remote job to fail with a segmentation violation. This happened after the jobs were received in the queue but before they were printed. When a new local print job was received, this would flush the queue, so all jobs would be printed. The local job followed a path through the code which did not have the bug. -------------cut------------------------- This bug still appears in the code for lpr-0.33-1.i386.rpm. I reference a patch to lpr-0.33-1.i386.rpm which fixes the problem (small buffer sizes which are used to store queue file names). Please consider adding this patch to further lpr releases (or at least tell me why you don't if you decide not to). I found this problem when using an upgraded version of named (version 8). Using this version of named, the file input file names from a remote lpd seem to include the full DNS name of the remote server, thus pushing them over the 40 byte buffer size used in lpd.
can you mail me the patch? That webserver doesn't seem to be responding...
Thanks! Fixed in lpr-0.35-1.