I'm considering replacing our use of the system logrotate with this
module and have a concern. I believe the correct sequence is;
1. *File::Copy::move* the active file to its new location;
2. execute post steps (as this is where you'd HUP apache etc.)
3. compress the newly rotated file.
If (as at the moment) the file is copied, then for a process that
retains a handle to its log file (like apache), in the delay between the
copy finishing and the post steps executing, the application may well
write to the log file, that you then delete. As moving is an atomic
operation within the same filesystem on Unix, the application will still
have a handle to the newly rotated file and will continue to write there
until it is HUPped. Then you need to wait for all processes to
relinquich their handles to it, before compressing it.
If you agree with this summary, I can create and provide a patch to
achieve these changes. (Tho the wait may just be a sleep 10 ;-
RT for rt.cpan.org