Giorgos Keramidas wrote:

>On 2007-10-28 12:54, Andrew Lankford wrote:
>>Thbbt! I'm reading the catman version of MAKEDEV. Wish I could
>>disable that "feature". Oh well. I'll delete it all and rebuild it
>>again if needed. Thanks!

>Ah, that's it then
>My usual `installworld' steps include this too:
> # cd /usr/share/man
> # find cat[0-9] \! -type d -exec rm {} +
>This takes care of deleting any `stale' preformatted manpages, so the
>next time I ask for a manpage, it's going to be reformatted.

Seems to me that there must a logic error in man. Man makes the effort
to check to see if the unformatted manpage is newer than the cat-ed
manpage, and if it is, it will try to re-create the cat-ed manpage, and
will show you an nroff'ed version of the unformatted manpage if it
cannot recreate the cat-ed version.

Clearly that's breaking in this scenario, because the mtime of the new
unformatted manpage, while presumably newer than that of the old
unformatted man page is still older than the mtime of the cat-ed page.

What seems to be missing, is that when the cat-ed manpage is first
created, it should have its mtime set to the same as the mtime of the
unformatted manpage, not the time at which you happened to read the man
page. I don't see any evidence of that is the sources on my system
(which is quite old, but it certainly looks like the bug persists).

A bit of stat and utimes jiggery pokery looks like it would solve this
problem. An else clause for " else if (rename(temp, cat_file) == -1)
{" in /usr/src/gnu/usr.bin/man/man/man.c would look like the place to do

(Alternatively, you could view it as a bug of the install/upgrade
process, since it could be argued that the mtime of the unformatted
manpage should be the time that the manpage was installed. Presumably
that isn't happening, else you wouldn't be seeing the cat-ed version).

Of course, with modern systems where nroff-ing a man page takes
negligible time and system resources, it could also be argued that
cat-ed man pages should be a thing of the past :-)


_______________________________________________ mailing list
To unsubscribe, send any mail to ""