Alessandro Vesely wrote:
Gordon Messmer wrote:
Sam Varshavchik wrote:
I think that ctlfile should be opened for append mode.
...capital idea. Am I wrong to assume that all of the calls to open()
in the openctl() function should have the same flags?
I'm not sure what you mean.
I mean that in the openctl function, there are three calls to open()
which, I would assume, should have the same flags. That wasn't true
originally, and I hoped that Sam would correct me if I was wrong to use
the same flags in all three instances.
And I wonder what sense may have
to truncate and append at the same time. (And O_TRUNC is not
needed on systems where O_CREAT | O_EXCL make the open fail
in case the file already exists.)
That's true, but the O_TRUNC option shouldn't break anything, so I
didn't bother to remove it.
[...]
- O_WRONLY | O_TRUNC | O_CREAT | O_EXCL,
+ O_WRONLY | O_TRUNC | O_CREAT | O_EXCL | O_APPEND,
PERMISSION);
An lseek(fd, 0, SEEK_END) upon return from the filter should suffice.
It should, but O_APPEND is easier and probably better in the sense that
every record written to a control file is supposed to be appended to the
end. If the file is opened with O_APPEND, that will always be the case.
I posted a patch that would seek after filtering. You can compare the
two approaches if you want to.
34 messages in net.sourceforge.lists.courier-usersRe: [courier-users] ctlfile
.patch
