On Thu, Aug 14, 2008 at 03:03:54PM +0200, Thibaut VARENE wrote:
> On Thu, Aug 14, 2008 at 9:24 AM, sword <[EMAIL PROTECTED]> wrote:
> ********************************************************************************
> > except for one case: a request for bare "/musicindex", but that does
> ********************************************************************************
>
> Waitaminute, that's a very good catch. I have the feeling the "fix" is
> very simple indeed: can you try replacing in musicindex.conf the
> following line:
>
> Alias /musicindex/ "/usr/share/mod_musicindex/"
>
> with
>
> Alias /musicindex "/usr/share/mod_musicindex"
A most excellent catch by you! It does indeed fix the problem. See
below for a consequent minor bug exposed and a quick suggestion to
wrap all of this up.
> Yeah but the subreq is actually being requested for "/musicindex"
> *without* the trailing "/"
Doh, I should have caught that as soon as I saw that edge case.
Sorry!
With your suggested Alias fix in place, it's now successfully reading
the Aliased image where it should be:
access("/usr/share/mod_musicindex//valid-xhtml11", R_OK) = 0
before generation of the page, but note the doubled slash.
During request time, everything is working OK:
read(7, "GET /musicindex/valid-xhtml11 HT"..., 8000)
stat64("/usr/share/mod_musicindex/valid-xhtml11") and open.
So (pardon the slow thinking),
subreq(/musicindex) is now Redirected to /musicindex/ by the Alias,
which is returning /usr/share/mod_musicindex/, which is then
strcat'ed together with /valid-xhtml11.
Looks like the right way to do this would be to have subreq look for
/musicindex/ and delete the leading slash in the access()ed filename
(i.e., change "/valid-xhtml11" to plain "valid-xhtml11". That should
work regardless of trailing slash or not in the Alias and gets rid of
the double slashitis at the same time.
(As a point of reference, grep -r Alias shows most of the other
packages do "Alias /doc/" with the slash, except for /manual. So, I
guess stick with convention and leave the Alias as shipped now and
twiddle the code. It's a small change either way.)
> Please let me know if the above suggested fix actually solves the
> problem, I'll quickly roll out a new package ASAP (I'm currently not
> able to do any testing unfortunately)
Thank YOU for your patience with me! Looks like you'll be able to
close up both bugs soon.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
