update-website.mk make script ready

Anderson Lizardo lizardo at linuxfromscratch.org
Wed Aug 11 11:18:20 PDT 2004


On Wednesday 11 August 2004 10:52, Jeremy Huntwork wrote:
> Fine by me.  Just one point that you may not have noticed.  Your script
> seems to exclude the online books.  When I ran my test syncs the books
> were included, now, however, you'll see error messages if you try to
> view:
>
> http://lfs.jenacon.net/view/stable
> http://lfs.jenacon.net/lfs/view/stable
> http://lfs.jenacon.net/blfs/view/stable
>
> Unless, of course, you intended to exclude them ;)

Yes, it was intended. Right after we start using the post-commit process, we 
should manually run "render-{lfs,blfs}-book.sh" to populate the content of 
these dirs, then next time, their own cron jobs will take care of this. I 
think we can call these scripts from the "svn-checkout" target (I should 
rename this), as it was just created to make an initial website tree.

I'd like to avoid doing book renderings (and possibly patches/hints updates) 
on our post-commit script, to make it as fast as possible. And, IMHO, it's 
more logical to have each project manipulate its files idependently. So, if 
the other teams decide to adopt a post-commit process too, the files they 
create on the website tree will only be updated (and thus rsync'ed) if they 
were really changed.

This will require some kind of orgnatization, mainly enabling a "lock" scheme 
that should suspend any updates to the rendered website tree right before the 
rsync iteration, and unlocked after the website is fully propagated between 
the mirrors. I think the "lock file" scheme we use currently to avoid 
multiple instances of the update script can serve here, we just need to make 
it global to any scripts that touch /home/httpd/www.linuxfromscratch.org/.

-- 
Anderson Lizardo
lizardo at linuxfromscratch.org
http://www.linuxfromscratch.org/



More information about the website mailing list