Before any upload is submitted the upload is registered in a list of
known uploads; this is needed to retrieve the upload object at several
points of the upload process. When a chunked upload is submitted first a
directory to upload all the chunks is created and, once that is done,
the chunks are sent; in order to send a chunk the upload object needs to
be retrieved from the list of known uploads.
When all the active uploads were finished the list of known uploads was
cleared. However, an upload is not active until it actually starts
sending the data, so while waiting for the upload directory to be
created the upload is already in the list of known uploads yet not
active. Due to all this, if the active uploads finished while another
pending upload was waiting for the upload directory to be created that
pending upload would be removed from the list of known uploads too, and
once the directory was created and thus the chunks were sent a field of
a null upload object would be accessed thus causing a failure.
Instead of removing all the known uploads at once when the active
uploads finish now each upload is explicitly removed when it finishes.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
The jQuery File Upload plugin triggers the "stop" event once there are
no more files being uploaded (even if some of them were added when
another upload was already in progress). Therefore, the progress bar
should be hidden in the "fileuploadstop" callback.
In some cases the "stop" event is not triggered and thus the progress
bar is not hidden once no more files are being uploaded. This is caused
by a race condition and it will be fixed in another commit; except in
buggy cases like that one (that need to be fixed anyway) it is safe to
hide the progress bar in the "fileuploadstop" callback.
In any case, note that the callbacks in "fileuploaddone" may be called
after the "stop" event was triggered and handled when using chunked
uploads. In that case once all the chunks are uploaded the assembled
file is moved to its final destination, so its promise could be resolved
after the "stop" event was triggered. Therefore a different approach
would be needed to keep the progress bar visible until the chunked
upload is truly finished, but for the time being the current one is good
enough.
Before this commit the progress bar was being hidden when the first
upload finished, either successfully or with an error, no matter if
there were other files being uploaded too.
The progress bar was being explicitly hidden also when the upload was
cancelled. When an upload is cancelled all the single uploads are
aborted, which triggers a "fail" event for each of them. However, the
"stop" event is always triggered when no more files are being uploaded,
so it is triggered too once all the single uploads were aborted. As all
the single uploads are immediately aborted in a loop when the general
upload is cancelled it makes no difference to hide the progress bar when
the first single upload is aborted or when all the single uploads were
aborted, so the progress bar is no longer explicitly hidden in the
former case.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
There were two routes apps/files/ajax/download.php but apparently also
apps/files/download.php I could not find any use of it.
Signed-off-by: Roeland Jago Douma <roeland@famdouma.nl>
The div that contains the elements related to the creation of new files,
and thus the upload button, is always present in the DOM; it is hidden
or shown based on the folder permissions by adding or removing the
"hidden" CSS class. However, as the other CSS classes for the div are
"actions" and "creatable" and a "display: flex" rule was defined for
".actions.creatable" below the "display: none" rule for
".actions.hidden" the last one took precedence and the div ended being
always visible, even if the "hidden" CSS class was set. Now the rules
for the ".actions.hidden" selector are defined below the rules for the
".actions.creatable" selector and thus the "display: none" rule is
applied as expected.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
Menu and home are not always visible; home is always visible, but menu
is shown only when needed.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
"getTotalWidth" is not more accurate; it is simply not clamped.
Moreover, "width/outerWidth" could be used in tests too, and also even
if "getTotalWidth" could be used in tests while others not that would
not be something to be stated in the API documentation, but in a
comment.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
After the changes in the previous commit "_showCrumb" no longer shows
the menu, only the same crumb that was hidden by the last call to
"_hideCrumb". Therefore, if the crumb was hidden because it did not fit
there is no need to try to show it again, as it will still not fit.
Moreover, the calculated width for a hidden element is not always
accurate; in some cases the calculated width is lower than the actual
width (it happens, for example, when using a background image like the
"Share" icon), which causea the crumb to be shown even if there is not
enough room, which in the end causes the siblings to overflow the
contents.
No unit tests for this one, though; you will have to trust me on this,
sorry ;-)
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
The crumb for the menu was shown like any other crumb when calling
"_showCrumb", but it was also shown when other crumbs were hidden
without taking into account the available width. This caused several
related problems, like the breadcrumbs taking too much space when the
menu was sometimes shown after the rest of the crumbs were adjusted to
the available width, or the menu being shown instead of the last crumb
even if there was room for it when the available width was increased.
Now the menu is always hidden before starting the resizing of the crumbs
to ensure that whether it was previously shown or not does not affect
the result. In a similar way, the menu will no longer be shown by
"_showCrumb", as it is not a regular crumb that has to be shown simply
if there is enough room. The menu is now shown as soon as any other
crumb is hidden; this ensures that the menu width will be taken into
account in further width checks. As when _updateMenu" is called it no
longer needs to take care of showing the menu this fixes the issue
revealed when fixing the test setup in the previous commit.
Finally, this implicitly fixes the failure in the breadcrumbs tests when
run on Firefox, as it was caused by the menu interfering in the
calculations of the other crumbs when increasing the width.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
The "Shows only items not in the breadcrumb" test was failing when run
on Firefox, but not on PhantomJS. This was caused by the differences in
the starting width between both browsers and an incorrect setup of the
test (the width set for the crumbs was overriden when the breadcrumbs
were rendered again, and the breadcrumb was resized to 300 from an
indeterminate initial width).
Now the crumbs are rendered and then its width, padding and margin are
set to a known value. Then it is resized to 1000px, which ensures that
there will be enough room for all the crumbs and thus the menu will be
hidden, and finally it is resized to 300, which causes the middle crumb
to be hidden and the menu to be shown.
Note, however, that the test now always fails, no matter if it is run on
PhantomJS or on Firefox; if the menu crumb is hidden when "_updateMenu"
is called it will show it, but it will also wrongly try to add the menu
itself to the menu. As the "crumb-id" of the menu crumb is "-1" this
causes the last regular crumb to be added to the menu. This will be
fixed with other related issues in the next commit.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
When calculating the total width of the crumbs only its padding was
taken into account; now the margin is too. In a similar way, before
showing a crumb only its width was taken into account; now its padding
and margin are taken into account too.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
This ensures that the resize tests do not depend on the values set in
the CSS files.
Note that this change causes a test to fail with Firefox, but not with
PhantomJS. This is due to a difference in the starting width used by
Firefox and by PhantomJS, and it will be fixed in a following commit.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
When the parent element of the breadcrumbs was resized to a larger width
and the siblings of the breadcrumbs expanded to fill all the available
width some crumbs could be hidden even if there was enough room for
them. The reason was that the width of the siblings being used to
calculate the available width for the breadcrumbs was the expanded width
of the siblings. Now as many crumbs as possible (that is, fitting in the
parent, no matter the siblings) are first shown so the expanding
siblings are compressed before calculating the available width.
Due to the lack of support for flexboxes in PhantomJS the related unit
test is skipped; it has to be run in other browser, like Firefox.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
Other apps could add elements to the controls outside the creatable
actions div (for example, the button to switch to the gallery), so the
widths of all the visible siblings of the breadcrumbs have to be taken
into account in the size calculations.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
There are some differences in width handling between the browsers used
to run the tests, most likely due to their support (or lack of) of
certain CSS features: PhantomJS requires "width" to be set (probably
because it does not handle flex displays and treats it like a block, so
"min-width" does not matter in this case), while Firefox requires
"min-width" to be set (otherwise the children of "#controls" could be
compressed due to its use of flex display and the elements would end
with a different width than the one needed for the tests). Due to all
that the width of the breadcrumb siblings must be specified in the tests
using both "width" and "min-width".
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
There is no need to call "setDirectory" again in resize tests; it is
enough to simply resize them (and isolates them better to just test the
resizing behaviour).
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
The "usedWidth" attribute was not used elsewhere outside the "_resize"
method, so it was replaced with a local variable. Moreover, it was also
renamed to a more suitable name ("availableWidth").
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
Setting the width of the parent element of the breadcrumbs and then
explicitly calling "_resize" is enough to test the resizing behaviour.
This makes possible to remove the "setMaxWidth" method and its related
code, which was used only for testing purposes.
Signed-off-by: Daniel Calviño Sánchez <danxuliu@gmail.com>
