Rendering Subversion version of HEAD

Jeremy Utley jeremy at jutley.org
Wed Jun 30 14:18:51 PDT 2004


Matthew Burgess said:
> On Wed, 30 Jun 2004 13:28:03 -0700 (PDT)
> "Jeremy Utley" <jeremy at jutley.org> wrote:
>
>> Something to be aware of when trying to set up render-on-commit -
>> whatever commands are used to render will run as the user who *MAKES*
>> the commit, so all LFS editors will need to have write access to the
>> directories where the HTML is written to.
>
> If we make the httpd user a member of the svnlfs group and make those
> directories group writable I think that'd work, but is it secure enough?
>  Additionally, render-lfs.sh will have to be group executable of course.

No.  The commit scripts will run as the user MAKING the commit, not as
user httpd.  So render-lfs.sh will need to be group-executable by one of
the groups that has LFS Editors, and the lfs/view/unstable directory on
the server will need to be owned by that same group, and also
group-writeable.  Otherwise, the render WILL fail - I ran into this when I
did render-on-commit on my server for BLFS.

>
>> Another thing we MIGHT want to consider is adapting to run "make
>> validate" first, to ensure the commit is sane - if it isn't sane,
>> write an error back to the person doing the commit, and NOT render the
>> book again.
>
> Yep, I was gonna do the pre-commit hook at the same time as the
> post-commit hook.  Only one set of tests to carry out then :)
>
Perfect.

-J-




More information about the lfs-dev mailing list