[Debichem-devel] [Debichem-commits] r512 - in /wnpp/gabedit/debian: gabedit.1 gabedit.1.xml rules

[Daniel Leidert]

Am Donnerstag, den 14.06.2007, 07:10 +0800 schrieb LI Daobing:

> suggestion on gabedit.1
> 
> 1. if you choose to ship gabedit.1 in debian directory, you should not
> regenerate it in build process, you can keep the generatrion rules in
> debian/rules 

It should not be regenerated during "normal" operation, like package
building. The chain only "regenerates" the manpage, if (a)
debian/gabedit.1 has been removed (that's the reason, why I will refuse
your second suggestion to avoid rebuilding the manpage every time) or
(b) if the timestamp of the XML source file is newer than the one of the
generated manpage.

But how should this happen? The SVN will always contain the manpage. If
changes to the source are done, running `fakeroot debian/rules
debian/gabedit.1' automatically updates the manpage. Try to run it again
and make will tell you, it's already up-to-date, so no further action is
taken. And this should also be the case during a build run. The
build-target checks for the existence of debian/gabedit.1.
debian/gabedit.1 itself (as target) checks the timestamp of
debian/gabedit.1.xml. Now it should detect, that it's own timestamp is
newer than the one of the source and nothing happens. The manpage is not
regenerated. But the whole depenncy chain makes ure, that (a) the
manpage exists and (b) it's not older than it's source (so it's always
up-to-date).

Compared to the possibility, that I may forget to update the manpage
manually (and upload them to SVN) after doing a change to the manpage
source, the advantages are still on the side of the current solution.

We use this way for shipping the manpages with their XML sources (with
upstream) for e.g. gnome-chemistry-utils (docs/man/) and
bluefish-unstable (man/) if you want to take a look.

> but not let "build" depend on it.

This is only a dependency chain to make sure, the manpage exists and is
up-to-date.

> 2. or you can choose not to ship gabedit.1, then generate it when call
> "build", but you also need remove it when call "clean".

See above. That's IMHO further a waste of time. There is no need to do
this.


My build logs do not show any unwanted regeneration of the manpage
during an svn-buildpackage/pdebuild build. Does it happen for you? If
yes, would you be so kind to send me the output of

fakeroot debian/rules -d $target_that_regenerates_the_manpage

then? The -d option adds debugging output, that gives more information
about targets and dependencies.

Regards, Daniel