[Home]ToothyWikiInternals/ErrorReports

ec2-3-215-182-36.compute-1.amazonaws.com | ToothyWiki | ToothyWikiInternals | RecentChanges | Login | Advent calendar | Webcomic

Old error reports that do not need action moved to ToothyWikiInternals/FixedErrorReports :




RecentChanges missing


Not much more I need to say really. --Admiral
Ah, it's back. --Admiral
It has happened again. Amusingly, if I look for recent changes from a while back, it displays properly, but if I look for changes since a recent time, they are missing. --Admiral
What URL are you using? What are you actually seeing, and what do you expect to see? AFAICT everything is working correctly here. - MoonShadow
I'll let you know if it happens again, but it's likely it only happens when there is a single change, so it wouldn't be around for long. --Admiral


Can't re-use image slots


As requested on the [image server] page, I tried to re-use an image slot (192).  The new reported file size is correct, but the preview and image itself are still the 'blank' image. --M-A, 8/4/07
I see the correct image here. Could your browser have been caching it? - MoonShadow
It does seem fine now.  Yes, I'll blame the cache, although I was trying F5 - maybe Ctrl-F5 was needed. --M-A



Backslashes at the end of lines appears to be handled really quite weirdly. --Admiral

Flibble  Flobble

Flibble  Flooble

Flibble \ Flooble

Flibble \
Flooble

(MoonShadow) A backslash at the end of the line, followed by zero or more units of whitespace, will *always* get eaten and turned into a single whitespace, resulting in the current line being joined to the next one (UseModWiki 0.92 formatting convention). In addition, two adjacent spaces become   followed by a space (request by AlexChurchill, don't remember the reason but it involved ASCII art) and any newlines left over become BR tags (same request). The examples you provide are consistent with this behaviour. If you really really don't want your backslash to be eaten, you can escape it as an HTML entity, or stick a nonbreaking space after it:
Fair enough - it just took a little working out. Now I know how to do it, it's not a problem. The behaviour I expected was to be able to escape a backslash (\\), and for that to override the end of line stuff. --Admiral
I thought it was me, asking for correct behaviour for spaces after full-stops.  Like that.  --Vitenka
Mph. You might be right. It was a long time ago, and I have no memory. ^^; - MoonShadow, lying scum

a \
b

a \ 
b

If people want this behaviour changed, suggest alternatives here ;)
Uhh.  Do we want line-continuation at all?  I can't think of anywhere it has been used.  Whilst it makes sense from a 'standard c usage of backslash' POV, I'm not sure how it fits in with the 'type as intuitively as you can' theme.  --Vitenka
It's been used somewhere. Not sure where though. --CH
I can't think of anywhere offhand, actually, and I can think of places where it's got in the way, though I'm willing to be demonstrated wrong by example :) Votes - can anyone see a good reason for having it, or shall I just chop it? One consideration is being able to type really long bullet points - does anyone's browser have a line length limit they're aware of. - MoonShadow
No apparent limit here.  I vote for cutting it.  But I bet I end up needing it as soon as it's gone.  --Vitenka
It gets used for discussion inside numbered lists without resetting the numbering. See MagicTheGathering/ScavengerHuntList. The idiom is:
  1. number
    comment?
    comment2
  2. number2.
It wasn't me that requested the preserve-multiple-spaces thing (I thought it was M-A), but I use it and like it. --AlexChurchill

Ohhh!  Is that how you do it?  Thanks!  I'd prefer a less clunky 'hash-but-don't-restart-counting' symbol, but if this is the only option, then I change my vote :)  --Vitenka

Fair - it stays ^-^ - MoonShadow



Page WebComics - text Demons/Daemons?  - link reifies to WebComics/Daemons?.  This is nonintuitive to say the least.

Not (easily) fixable. Demons is not a valid page name (valid page names must have at least two capital letters separated by at least one lowercase letter) as far as the wiki is concerned, so /Daemons?, which *is* a valid subpage name, cannot be a subpage of Demons unless it is forced to be by double square brackets. By the time link processing happens, information on preceding text has been lost. Backtracking within the regexp to exclude classes of preceding characters is too expensive. I could make the behaviour slightly more intuitive by making subpage names adhere to the same strict naming convention as main page names currently do, if people want. This would break a moderately large number of links currently in use. I don't want to relax the main page naming convention to match the subpage convention since *all* capitalised words would then become wiki page names. People could put spaces around their /'s if they don't mean to write a wiki page name. Any other suggestions?




Typing ISBN in upper case becomes a link without a colon.  --Vitenka

It wasn't a bug, it was *meant* to work like that. But you're right, it's not very intuitive. So I've changed it. - MoonShadow




Mixing # and * lists doesn't work:
  1. This
    1. works
    2. as expected


That's not a bug report, that's a feature request ^^; - MoonShadow
Really?  I just assumed that lists would work completely, since they seem to work in the usual case.  Guess I'll change it to all *'s.  --Vitenka

(MoonShadow) Lists are currently generated in a single pass by code like this:
 .
.
} elsif (s/^(\*+)/<li>/) {
(generate UL, push /UL onto HTML stack)
 } elsif (s/^(\#+)/<li>/) {
(generate OL, push /OL onto HTML stack)
 .
.


I sometimes, when following the (diff) link in RecentChanges, get a backwards diff, which is odd. I haven't worked out what causes it, but it seems to be 10-20% of all pages, and to happen repeatably for any one page, for a given edit. Is this in any way deliberate? --TI
Yes, this is deliberate - it is the simplest way of making the edit links at the bottom of a RecentChanges diff point to intuitive places. See discussion near the bottom of ToothyWikiInternals/WikiFeaturesDONE. - MoonShadow

I believe the reason that you sometimes get the backwards diff is tied to the fact that the dif links on RecentChanges now use numbered revisions and that, while the current dif is displayed forward, all past difs are displayed backwards.  Hence, if you haven't refreshed RecentChanges in a couple of minutes and someone makes another edit to a page, you will get the dif you expected, but backwards, instead of the most recent dif.  I'm still trying to decide if I like it this way or not.  It can be a bit confusing at times.

The reason is that the original wiki code's "edit this page" link would cause the revision currently being displayed to get edited when clicked on. This is perfectly intuitive if you are deliberately looking at a revision other than the current one, but counterintuitive if you clicked on a RecentChanges diff-between-two-revisions (not a feature the original code supported) and expect to be able to add your own comment to the current version from the page you then see. It was rendered obsolete by my adding explicit links for both actions and making it harder to edit old versions of pages, so I shall probably be getting rid of it soon. - MoonShadow

One upshot of this is that if you view the most recent diff, wait a while (until someone else has made a revision) and hit refresh, the diff will suddenly reverse direction from forward to backwards.  In the old behaviour, it would have displayed the new latest diff instead.  Hope this helps, its a little confusing but probably unavoidable --K

This is an inevitable side-effect of tying each diff link on RecentChanges to a particular edit or set of edits (and therefore to a particular pair of revisions) - which was a feature requested by *lots* of people.. - MoonShadow

ChrisHowlett is not convinced that this is the whole story - or at least is partially unsatisfied by one consequence (I can cope with the odd diff going backwards instead of forwards), and one thing that might be a consequence. Firstly, if the last change to a page was a minor edit, (diff) links on RecentChanges and the ticker links show a backwards revision from last major to previous revision, display the last major revision and therefore miss out the minor edit - see ChrisHowlett/Quotes as linked from RecentChanges (WikiNow 14/5/05, 18:22). Secondly, ticker links often show backwards revisions of great length - I've seen Revision 50 to Revision 1, and others. Not all are diffs with 1, I don't think. This seems to be most prevalent on "hot" pages, so that the ticker is one or more edits behind the wiki.
The RSS feed script generates diff links to cover changes made over the entire time period (24 hours if you're using the default settings); hence the large timespan. This behaviour is deliberate, but can be changed if people really don't like it. I ought to do something about the other thing you mention. - MoonShadow
But the ticker link for God/MultiTheist currently gives just the single major diff by SF. Looking at the link, this actually bemuses me slightly, as it's http://www.toothycat.net/wiki/wiki.pl?action=browse&diff=1&id=God/MultiTheist&diffrevisionwith=32&revision=57 - is this an argument-order problem, perhaps? I note in passing that revision 32 was not the closest one to 24 hours ago by a long chalk. --CH
...and that diff link now does what it would be expected to, since Requiem made another change. I'm so confuzzled....
Requiem apologises for editing pages in what he knows to be a highly confusing fashion. But he often clicks 'Save' before he means to.
Ah, indeed. The most recent edit is now minor, whereas it was major previously.


The UnreifiedLinks list includes things like "InABubble_Go" (from AngelasBoot).  These are misgenerated from the page contents. --M-A

Hm. Probably a problem with my indexing script - I'll look into it. Thanks ^^ - MoonShadow


As demonstrated at the top of MagicTheGathering/EventsAndReports, the $$CONTENTS directive stops its links to the relevant subheading when it reaches a Wiki link, but doesn't start it again after the link stops. With subheadings which start with a link, this means one has no way to jump straight to the subheading. A fair option might be to remove links before making the subheading contents text, as the link will always still be available down at the actual subhead itself. --AlexChurchill
Um - that's not actually trivial at the moment; by the time I know I'm gonna make something into a heading and therefore know it'll need adding to a contents table, it's already had the common markup performed on it - for quite a long time, even - which includes wiki links. I'll have to have a long think about quite how to solve it without breaking all perl-line markup. - MoonShadow


Diff links appear still not to work correctly (same problems as above). To clarify:
Link format tweaked. Please let me know if this has helped or broken anything. (The links in your post above will still exhibit bugs, test ones currently in the ticker or in RecentChanges) - MoonShadow
Much better, thanks! A quick look seems to show it working exactly as would be hoped - diffs purely for minor edits do show the minor edit being changed, and go forwards. Collated ticker-diffs (ie, ones made with the ?s option) do indeed show the whole day's-worth. The only thing I would suggest (and I realise it's a feature request rather than a bug report) is that it would be nice if RecentChanges diffs when you're not showing minor edits diffed from last major revision to the current one, rather than (as currently) last major revision to the one just following, displaying the current revision as text. I can probably clarify that if it's a little unclear. --CH
Hmm, interesting. As a habitual ticker-user to the point where I only normally visit RecentChanges itself when using other people's computers, I'm not convinced I like the change. The new version shows a whole day's worth of changes, but seeing as I've typically seen every change except the most recent one before anyway, I tend to just want a link to the most recent diff. Is there a parameter tweak I can make to my ticker feed to recover links to just the most recent diff? --AC, apologising for causing trouble
To just show the last change, drop the "s" from the query string your ticker uses. There's currently no way to list all the changes but only provide a diff link to the last one - is that what you want? - MoonShadow
Hmm. Am trying it out without the s for the first time, and I think I like it. It'd probably crowd over quickly on busy days; maybe I might actually use folding. It does have the entertaining consequence that today's ticker proudly proclaims "Spam[*] Spam[*] Spam[*]". My ideal would be a feed that combines, on one hand, the "s" behaviour of combining all edits to the same page into one link with all edit comments in the [*] ; with on the other hand the non-"s" behaviour that the link when clicked takes you to the most recent diff. This is pretty close to the way the behaviour used to be, which I think I preferred; however, I'm still considering the non-"s" behaviour. --AC
*frantic typing* OK, try sr. - MoonShadow
I thought the accepted transcription was clickety clickety?  --Vitenka
PeterTaylor thinks that has certain connotations which aren't necessarily appropriate in this context.
I'm not going near the ticker, just in case :)  --Vitenka
Yay, brilliant. Thanks! :) --AC




Am I just forgetful and we discussed this, or is clicking 'diff' on a minor edit now supposed to show you JUST that minor edit?  I was used to it showing you everything down to the previous major edit.  --Vitenka
Hm. Looking at the script, the original UseModWiki code would show you the difference between the latest revision at the time you click on the diff link (i.e. if someone else has edited the page since the RecentChanges you're looking at was generated, that's the latest revision) and the previous major, minor or author revision depending on your preferences. Current code saves to the recent changes log file the current and previous revision ID each time the page is changed, and the RecentChanges diff link goes to a diff between those. I do not remember why the behaviour is like this or who asked for it. I didn't even remember that this is what it was doing :) - MoonShadow
This changed relatively recently, I believe at the same time as when the diffs returned by the ticker changed. I liked major diff being the default (with other diffs available in brackets). --AlexChurchill





Double BlueQuestionMark:  --SunTsu??
Presumably caused by the -- making it a signature link in addition to an unreified link. Not that that's desirable behaviour; just noting the probable cause. --CH
That's been known for ages. It's what happens if you signoff with a proper CamelCase? WikiWord that's also unreified, after a double dash. --AC
Indeed, ItsNotABugItsAFeature. The fact that the page you are using as signature is unreified should be a big fat warning sign. In this case, the reified page is called SunTzu. - MoonShadow
Fair enough - the spelling was kept different in this case deliberately.  --Vitenka
Oops.  Sorry, I corrected before reading this.  Why did you want it different? --K
Doesn't matter - but it was atytributed to the forum poster I knicked it from.  Definately not the original.  --Vitenka




Very minor bug: link to chess gets square brackets around it when (I don't think) it should (link to sandbox doesn't).  Presumably due to the hyphen in the page name? --K
Indeed it was. Now fixed. - MoonShadow

Um.. is it possible something else is now broken? The links at the top of the Artifaxis comments pages now don't work - they've just got little blue question marks beside them.. - SunKitten
Got it. Not sure *why* that used to work as written, but fixed now. - MoonShadow




Diff with very small change seems to be wrong. Have a look at http://www.toothycat.net/wiki/wiki.pl?id=PenroseTiles&action=browse&diff=1&diffrevision=1&diffrevisionwith=2

I added a space in the middle of a word, in the the middle of a link. --Admiral




I tried to post a link on Nat/PresentIdeas from a small model shop and got a scary message back saying my IP address had been blacklisted. I think the Wiki objected to the fact that its site is hosted on freewebs. What do I do in this situation. Can I not link to it? Google "c-rail intermodal" to see the site. It has some kind of proxy address, but I was trying to link to a specific page. --Nat
Probably an overzealous regexp on my part. The Google search gives [this], which isn't blocked - I assume that's not what you meant? Post the problem URL, less its http://, and I'll unblock or whitelist it depending on quite what rule it's triggering. - MoonShadow
Thanks MS.  Here's the link I'd like to post: http://www.freewebs.com/crail=intermodal/4mm.html Please let me know when it's safe to put this on the other page.  Thanks, Nat.
OK, unblocked. - MoonShadow

Trying to save a revision of XComEnemyUnknown has similar problems, because www.the-underdogs.info/ is blacklisted. --SF
Whitelisted. - MoonShadow


Backlink searches seem to be just turning up two or three pages, even when hundreds should result - for example backlink [on StuartFraser] only produces one page. --SF
Fixed. Sorry. - MoonShadow


IBSN links pull in only the next 10 digits. Should you be supporting ISBN-13? --CH
I've not seen Amazon use ISBN-13 links, so not sure how to support them. Currently their ASIN stock codes appear to be a superset of ISBN-10. I guess I could recognise ISBN-13s in 978-space and spit out the appropriate ISBN-10 links. - MoonShadow
Fair enough, no vast point in support then. I'm sure you know this already, but worth noting that the ISBN-10 number I just gave differs from the ISBN-13 suffix in the final digit. --CH
Yes, the checksum calculation is different for ISBN-10 and ISBN-13. - MoonShadow

ec2-3-215-182-36.compute-1.amazonaws.com | ToothyWiki | ToothyWikiInternals | RecentChanges | Login | Advent calendar | Webcomic
Edit this page | View other revisions | Recently used referrers
Last edited June 8, 2009 5:24 pm (viewing revision 189, which is the newest) (diff)
Search: