collect v2.0 has some problems, some of which I have fixed.

The patches fix the following problems:

• sleeping for indeterminate times on heavily loaded system
• -l flag failed for large configurations
• system and user times in proc subsystem could be wildly wrong
• tape stats were zero
• process user and system time resolution (centiseconds) was too large. changed to microseconds
• PIDs on V5 can be 6 digits
• error message "scanf failed" at startup. (due to HSG "IDENTIFIER")

You can find the patched V5 sources here.

I updated this archive on 27-NOV-2000. In addition to the fixes above, this version contains a better fix for the "-l" (last-record-flag) problem. It can now determine when overflow of the record-length datatype in the mini-header has occurred and prints a warning and quits. It also advises you to use collect to 'filter' the binary collect file causing the trouble using collect -p <old-binary-file> -f <new-binary-file>. This causes collect to unpack each record and then repack it, which has the side-effect that the mini-header record-length is re-written, this time to a datatype large enough to hold it.

I haven't had time to integrate the patches into the V4 sources yet. sorry.