GitLab

Bug #1448807

Change-Id: I300c4d45d6a9c0568339a49dd3eeedce7ad9a117
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Jinelle Foley-Barnes authored Apr 14, 2015 and Robert Lyon committed Apr 24, 2015

Change-Id: Id127f4256d9419a65b38a78e4f610bc5119901ed

This is a behat test for bug (Bug 1424512) so it is dependent of that fix.
Also depended on the General node being able to expand (Bug 1407854)

Change-Id: I07112335965e44d9e80370bc0b3856d1cf38184d

This test needs to be able to expand nodes.

Block titles needs to be able to be clicked somehow that's human
readable if possible.

Change-Id: Ib79a4916b72643836fa7e43e558671c3b97324ad

This is because the selenium gets confused by having two 'save'
buttons on the page, as the 'addblock' one gets hidden but still exists

Change-Id: Ieef1c583fa0f4ea5381ba931ab12f116556740cf
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Change-Id: Iee5110eb9ffb9bcb1a6dbbcdfa1b8a43891db93e
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

On the basis of my profiling and testing over the last week,
all of the scripts affected by this patch don't need their
raise_memory_limit calls. Ensure_sanity currently sets the
memory limit to 128MB, which has been seen to be more than
sufficient for each of these use cases.

Tests have involved at least 1000 records being imported in
each case, and sometimes as many as 20,000. I would have
liked to have tested with some really full profiles (eg
lots of pictures and content), but am satisfied that there's
enough margin to cover those cases. In any case, such files
are often handled using external apps and therefore won't
be counted toward PHP's memory use anyway. Finally, the
limit can easily be increased by the user if necessary for
a particular case.

Change-Id: Ifecc83fd47da51268bae6cbd6960735eb91f9403
Signed-off-by: Nigel Cunningham <nigelc@catalyst-au.net>

Add caching of results to ArtefactType::get_plugin_name.
In a test export of 1000 users via admin/users/bulkexport.php,
this patch reduced the number of calls to get_field from 78,000
to 26 and reduced the execution time from 115 seconds to 92 seconds.

Change-Id: Ie7b2e39089bed1c3cd4844748b21f1ffb251f8dd
Signed-off-by: Nigel Cunningham <nigelc@catalyst-au.net>

Bug 1446426. Relative URLs don't work if your Mahara wwwroot
is a subdirectory, e.g. http://example.com/mahara

Change-Id: Idfbb4c1a81191a41c25fc816eec784cd14df7637

I have moved the feature files that were outside of folders into folders.

Go to mahara/tests/behat/features to see the ones that have been moved. They were moved in attempt to
tidy it up and check they are no redundancys.

Change-Id: I63db763a359d2cc715e0a4a6128cf01e774ebbab

Robert Lyon authored Apr 16, 2015 and Aaron Wells committed Apr 17, 2015

Bug 1286935

Seeing as we check the url against FILTER_VALIDATE_URL and that only
site admins can add to the 'allowed iframe sources' that should be
enough without having to add the / to the end of the url.

Change-Id: I82e3623d3df2fa03012278d334994224c51a092e
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Was trying to access 'Pete Mc' but the link for first and last name
are separate on the admin/users/search.php page so we need to just
look for the first name 

Change-Id: I42778395f8c1b77cf2bb40333c1d3fd935802ecb
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Bug #1422232

Change-Id: Ia8fd7d074db3be027e1318a07d062a9ed1bb2ad8
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Do correct string/variable comparison

Change-Id: I98c5c1360891699e439108789b2015d7587222ca
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

By doing two things:

1) Getting the embedded SWF object to set the
 allowscriptaccess = "never" and allownetworking = "never"

2) By forcing a 'download file' link to actually download file
- this goes for all files now that don't have embedded=1
in their url.

I've done it this way, having the embedded item have extra url param
so that if a user tries to manipulate a url by removing params it
will default to force download.

I've merged the changes I'd done here https://reviews.mahara.org/#/c/3522/2


and I've also cleaned up places where the download=1 was used as that is
not needed now. Now if there are places where we need to embed rather
than download we add the embedded=1 to the url.

Change-Id: If5290a7c571d06d4178ef2ae5c4c09ed287403b4
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

* changes:
  Behat Test: Friend Request (Bug 1440908)
  The accepting friend notification contains full url (Bug #1440908)

Rather than the loading spinners

Change-Id: I051fedd6bb8ee5061960c5f6bb2a67fb5b72d9ad
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Discovered this problem when I fixed the HTML export - it was
exporting blocks with the loading spinner rather than actual block
content

Change-Id: Id9c790f7cc6058141fb98c6d984af188ae265d6c
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Bug 1444229: It was hacky to rely on the block content being empty
to signal the AJAX block loader. We should tell it to do so explicitly.

Change-Id: I9816c43c96ffed85282ac0d874fa5bfe1ca62e00

Bug 1443770. Note this solution doesn't work fully in IE10.
In that browser, the keyboard nav works, but clicking on the
thumbnail won't pop open the accessibility links.

The fix for that, is ultimately to change this from using
MochiKit to using JQuery.

Change-Id: I5f173f6b32a3a301c10262cd71838f78821b36dd