Mailing List Archive

Ira Abramov wrote:
> http://cliki.site.co.il was running fine for several moons under
> 1.3.0beta3. I decided it was time to upgrade once 1.3.2 was out, but I
> assumed wrong that a "cvs up" will give me the current release.

Only if you're checking out the REL1_3 branch. If you're in the HEAD
branch you'll receive the latest alpha code.

> instead
> it landed a 1.4 alpha in my directory, and I got errors instead of
> pages. I then tried running the install process once more (it did such a
> fine job upgrading my DB from 1.2.6 to 1.3.0beta2) and no go.

Can you specify what the problem was? Without details such as error
messages, we may never be able to fix your problem.

> I assumed
> the DB itself was not ruined since nothing got converted, and simply
> untared a fresh 1.3.2 tarball on my site, reinstated my wiki.png and to
> my surprise the site was showing mostly ok, but some of the [[links]]
> are not parsed. as anyone can see on the site (even if you can't read
> Hebrew).
>
> I edit the page, the unparsed plaintext looks OK, but saving it gives
> the same results, some of the [[wiki links]] work and some are printed
> as is. I could not even find a common feature of the unparsed links to
> explain this.

First, make sure you really have a clean copy, that you haven't mixed
together files from two versions.

Second, try to isolate individual links on separate pages. See if they
are consistent. Try re-typing them and cut-and-paste. Check for
non-breaking spaces and other odd characters.

I would try this myself, but your wiki won't let me edit so you'll have
to diagnose this yourself.

-- brion vibber (brion @ pobox.com)

Quoting Brion Vibber, from the post of Wed, 01 Sep:
> Only if you're checking out the REL1_3 branch. If you're in the HEAD
> branch you'll receive the latest alpha code.

noted for next upgrade. my CVS skills were not refreshed often lately.

>
> >instead
> >it landed a 1.4 alpha in my directory, and I got errors instead of
> >pages. I then tried running the install process once more (it did such a
> >fine job upgrading my DB from 1.2.6 to 1.3.0beta2) and no go.
>
> Can you specify what the problem was? Without details such as error
> messages, we may never be able to fix your problem.

well, not without upgrading to it again.

I don't mind running 1.4 alpha and report bugs, but I could not find a
way to upgrade without diving into the code (like a doc telling me about
what changes to do to the table schemes), so I folded up the directory
and reinstalled 1.3.2 from the tarball.

> First, make sure you really have a clean copy, that you haven't mixed
> together files from two versions.

done. still no help.

> Second, try to isolate individual links on separate pages. See if they
> are consistent. Try re-typing them and cut-and-paste. Check for
> non-breaking spaces and other odd characters.
>
> I would try this myself, but your wiki won't let me edit so you'll have
> to diagnose this yourself.

it won't let people edit without logging in. Just a minor spam bot
stopper. I opened anon editing for now.

I retyped a few links, no go. copied a paragraph to the sandbox and as
you can see the category setting is still not parsed as well as a word
at the end of the segment, exactly like in the original page.

http://cliki.site.co.il/%D7%95%D7%99%D7%A7%D7%99%D7%A4%D7%93%D7%99%D7%94:%D7%90%D7%A8%D7%92%D7%96_%D7%97%D7%95%D7%9C

below it I cut and pasted a bit of the main page as well. tried
seperating the [[links]] by whitespace from the neighbouring commas and
such, still no use.

Finally on the last section I just typed, copy-pasted and played around
with several [[links]] that worked or didn't. Results: the exact same ones
that work - work again (whether to existing pages or red links to non
existing ones), and those that didn't get parsed or were random strings
I invented to new pages did not get parsed into links at all, red or
blue. no matter if I retyped or copied them.

Other voodoos I noticed:
* when I sit the "random article" link (third on the navigation panel) I
sometimes get an error where the article does not exist like so:
http://cliki.site.co.il/%D7%A7%D7%A0%D7%99%D7%94
this is the text of "badtitletext" which may help with the above
phenomenon, these are links that used to exist, and they infact DO
appear in the list of pages "Randompage" is looking at, but can't be
displayed when I try to go there. the Allpages demonstrates this the
best:
http://cliki.site.co.il/%D7%9E%D7%99%D7%95%D7%97%D7%93:Allpages

note, the error says this page is not displayable from what I gather,
it does NOT say "this page is blank, please edit it" etc. like if I
point to a random string that was never a page.

* Also asking for Special:Categories leads me to an error page:
http://cliki.site.co.il/%D7%9E%D7%99%D7%95%D7%97%D7%93:Categories
the line in question from function formatResult( $skin, $result )
calls a function on $skin but complains it's a non-object... the plot
thickens.

probably a few other mysteries I'm yet to discover...

any/all help and pointer will be appreciated.

--
Assimilated by the Borg
Ira Abramov
http://ira.abramov.org/email/

Ira Abramov wrote:
>>Can you specify what the problem was? Without details such as error
>>messages, we may never be able to fix your problem.
>
> well, not without upgrading to it again.
>
> I don't mind running 1.4 alpha and report bugs, but I could not find a
> way to upgrade without diving into the code (like a doc telling me about
> what changes to do to the table schemes), so I folded up the directory
> and reinstalled 1.3.2 from the tarball.

Run the installer, point it at your existing database. Same as always.

> this is the text of "badtitletext" which may help with the above
> phenomenon, these are links that used to exist, and they infact DO
> appear in the list of pages "Randompage" is looking at, but can't be
> displayed when I try to go there. the Allpages demonstrates this the
> best:
> http://cliki.site.co.il/%D7%9E%D7%99%D7%95%D7%97%D7%93:Allpages

All the problem titles contain the byte %A0; there seems to be something
wrong with the bit that strips out non-breaking spaces from titles. I've
checked in a fix, you can make the change manually. In Title.php in the
function secureAndSplit() change this:

global $wgInputEncoding;

to this:
global $wgUseLatin1;

and this:

# Disallow non-breaking space, except from interwiki links
if ( $wgInputEncoding == 'utf-8' ) {
$rxTc = '/\\xc2\\xa0|[^' . Title::legalChars() . ']/';
} else {
$rxTc = "/[^" . Title::legalChars() . "]|\\xa0/";
}

to this:

# Disallow non-breaking space, except from interwiki links
if ( $wgUseLatin1 ) {
$rxTc = "/[^" . Title::legalChars() . "]|\\xa0/";
} else {
$rxTc = '/\\xc2\\xa0|[^' . Title::legalChars() . ']/';
}

-- brion vibber (brion @ pobox.com)

Quoting Brion Vibber, from the post of Fri, 03 Sep:
> >I don't mind running 1.4 alpha and report bugs, but I could not find a
> >way to upgrade without diving into the code (like a doc telling me about
> >what changes to do to the table schemes), so I folded up the directory
> >and reinstalled 1.3.2 from the tarball.
>
> Run the installer, point it at your existing database. Same as always.

well, I think I did just that, but without first removing
LocalSettings.php :-)

oops...

>
> > this is the text of "badtitletext" which may help with the above
> > phenomenon, these are links that used to exist, and they infact DO
> > appear in the list of pages "Randompage" is looking at, but can't be
> > displayed when I try to go there. the Allpages demonstrates this the
> > best:
> > http://cliki.site.co.il/%D7%9E%D7%99%D7%95%D7%97%D7%93:Allpages
>
> All the problem titles contain the byte %A0; there seems to be something
> wrong with the bit that strips out non-breaking spaces from titles. I've
> checked in a fix, you can make the change manually.

<bow> four lines changed, apache restarted (otherwise mmcache or
whatever didn't see the change) and I'm back on the air... thanks!!!

I understand he.wikipedia is not running this version yet, or the
breakage would have been noticed earlier. I take off the virtual hat to
your amazing almost-telepathic debugging skills, I feel like I have met
the modern Sherlock Holmes :)

Have a good weekend, I'm off to updae my site :)

--
Cybersitter
Ira Abramov
http://ira.abramov.org/email/

Ira Abramov wrote:
> I understand he.wikipedia is not running this version yet, or the
> breakage would have been noticed earlier.

Actually, some ass hacked in a temporary fix on the live servers without
checking it in or telling anybody.

-- brion vibber (brion @ pobox.com)