Once again, bitten by Perl. We used to print a cache "if ($Page{blocks}
and $Page{flags} and GetParam('cache', $UseCache) > 0)" -- but if we
have exactly one clean block, then flags will be "0" which is false. So
now we're testing for defined $Page{flags}.
There was a huge discussion with a lot of tension:
https://oddmuse.org/wiki/Revolutionary_Changes#ForgiveAndForget
And also the comments:
https://oddmuse.org/wiki/Comments_on_Revolutionary_Changes#ForgiveAndForget
But in the end, it is safer to have a history which is not broken.
Don't get it wrong, ForgiveAndForget is still a good thing, it's just not what
we should do *by default*.
If your wiki does benefit from ForgiveAndForget, then add this to your config:
$KeepDays = 14;
Although this change solves a couple of important problems, it does not address
new ones that arise because of no ForgiveAndForget. Namely it does not
resolve the problem of deleting stuff when you *really* have to do it. For
example, [[DMCA Extension]] (or similarly named extension with the same
purpose) should be developed. These problems existed for a long time, because
people were using “$KeepDays = 0” a lot. It is just that we started to accept
wikis with no ForgiveAndForget more thoroughly.
In other words, this commit is just part of the bigger change.
Why don't we set it to 5 years? Because then it will be a time bomb that will
be triggered unexpectedly. We should have a more predictable default value.
Modules are not loaded recursively, so we are free to use any directory inside
$ModuleDir. It is also where translations are located in the git repo.
Also, %library was renamed to %TranslationsLibrary (which is now "our"). This
is required for tests and for custom configuration.
Remember the problem with toc.pl when the whole page was *sometimes* not
utf8-decoded? There were some thoughts that it might be associated with
memory files, and it is correct. Although I was not able to narrow it down
last time, now I did (simply because this problem appeared elsewhere).
If you look at $output variable after utf8::decode with Devel::Peek, you
will see two variants of flags. This one looks good:
FLAGS = (PADMY,POK,pPOK,UTF8)
And this one is wrong:
FLAGS = (PADMY,POK,pPOK)
This problem is weird because it works inconsistently. Most of the time
you will get correct output, but sometimes it will be broken.
Someone has to golf it down to something short in order to submit perl
bug report. This, however, does not look like a simple task.
Current workaround is as stupid as it looks like, but it works.
Somehow assigning it to another variable solves the problem (which, by the
way, is similar to solving other perl string-related problems).
We no longer require an existing webserver running a wiki at
http://localhost/cgi-bin/wiki.pl. Instead, we're running our own
stuff/server.pl on a random port and use it for testing -- and we kill
it when we're done.
ReplaceAndDiff calls Replace, which loops over all pages. That's why we
don't need to call it from SearchTitleAndBody -- that makes our code
runs way too often.
“There are no comments, yet. Be the first to leave a comment!” – that's what
you will see when you preview your comment on an empty page.
Since the user is already doing it, there is no need to tell that. Also, it
may look like it is part of the preview.
We no longer do that (after this commit). In other words, the preview should
look exactly like the resulting page.
The original issue was that looking at all changes (action=rc all=1) the
resulting diff didn't always make sense if you clicked on the diff link.
It showed the difference between that revision and the current revision.
The PrintHtmlDiff sub was changed significantly to make it easier to
understand and to help fix this issue.
The drawback is that it now requires a new key in page and keep files:
lastmajorsummary. It goes with lastmajor and diff-major and records the
summary for that particular edit. As new changes will start recording
this new key, the change will slowly propagate in existing wikis.
Whenever you look at minor diffs, however, the existing summary key is
chosen. Plus, whenever you want to look at differences between
particular revisions, this is equivalent to looking at minor diffs. So
the only situation that is problematic is an edit history like the
following:
A - major change
B - major change (major diff, major summary, last major revision)
C - minor change
When looking at this page with diff=2, we want to show major diff, major
summary, last major revision. If B happened before this commit was
installed, the summary will be missing.
Changing everything to return objects is a worthy goal, but for now we have
taken enough destructive steps towards it. Therefore, this commit fixes the
problem in backwards compatible way (by adding one more parameter to the
signatures).
Note that this additional parameter is NOT a timestamp, it is a whole page
object. Which means that we are still moving towards our goal of using page
objects everywhere, this commit is just doing it in a backwards-compatible
way.
sub ParseData is fully backwards compatible. If some module runs it in list
context, then it will get listified hash like previously. New code should
always run it in scalar context though (everything in our code base
was changed according to that).
sub GetTextRevision is not backwards compatible (don't let “wantarray” usage
to confuse you). Most modules do not touch that subroutine, so we are probably
fine (modules from our git repo that do use were changed accordingly).
“EncodePage(%$page)” looks wrong. It seems like we should change it to accept
hash ref.
Trying to solve an issue: sometimes the test fails on Alex Daniel's
test server but never on Alex Schroeder's laptop. The output of Recent
Changes being tested has no rollback button for one of the page links.
Actually, the last six edits have no rollback button:
12:34 UTC (diff) MinorPage . . . . 127.0.0.1 – Rollback to 2015-09-01 12:34 UTC (minor)
12:34 UTC (diff) AnotherEvilPage . . . . 127.0.0.1 – Rollback to 2015-09-01 12:34 UTC (minor)
12:34 UTC (diff) OtherPage . . . . 127.0.0.1 – Rollback to 2015-09-01 12:34 UTC
12:34 UTC (diff) NicePage . . . . 127.0.0.1 – Rollback to 2015-09-01 12:34 UTC
12:34 UTC (diff) EvilPage . . . . 127.0.0.1 – Rollback to 2015-09-01 12:34 UTC
12:34 UTC (diff) MinorPage . . . . 127.0.0.1 – testerror (minor)
Note that this includes the "testerror" minor edit which is about to
be rolled back. Perhaps that's because this should hold in
RollbackPossible and it does not: $ts != $LastUpdate. $ts would be the
timestamp of the testerror edit and $LastUpdate would be the timestamp
of the rollback. I've added another 1s sleep between these two.
