GitLab

With this commit we cannot fix the segmenter anymore
(the property checked of the input fix-segmenter is not defined).
See next commit.

See #1978 and !3

Aurélien Béliard authored Jan 12, 2017 and Mikaël Salson committed Jan 12, 2017

It prevents the segmenter from moving up and down

See #1978

.data is not used for years. Replace it with .vidjil

The downloaded files are now the same are the ones served by the server to the client.
Before that the served files were the ones stored on disk but get_data and
get_analysis do some work to slightly modify the files.

See #2102

Following 42d7ed86 and #2101 we ensure that such a problem won't appear again
by checking in dateDiffMinMax whether the dates can be substracted or not.
When dates can't be substracted the result will be NaN, which can be tested
and an exception raised.

Tests are done to ensure the correct behaviour.

Timestamp should not be ``deleted'' as a date is expected for the timestamp.
dateDiffMinMax can also deal with a "None" string. So we change the string
output here, which is less specific, but which works correctly with what is
expected in model.js.

The name of the sample is still "deleted" which will help remember that this
file has been deleted.

Fixes #2101.

app moved, the repository did not. Until we have decided on a generic
name for the backend server, vda will have to suffice.

Although root should work, it's always safer to run things as non-root
users

Some values were still set to some of the values used during
development.

docker uses the vidjil packages which do not contain the git metadata

the fuse server was referenced as localhost in the code. However, for
the separation of reposibilities to work correctly in the docker
installation it was necessary to add a variable in order to be able to
reference the correct container for running custom fuse jobs

FS_LOCK_THRESHHOLD and SCHEDULER_HEARTBEAT were missing from defs

The repository address was set to a server-specific address. This one
should always point to the correct server

Some volumes were mounted in a different location from the rest of the
volumes.
Now everything will be easier to find in one single folder

The mysql volume was by default set to the hosts mysql installation.
Although it was convenient for some parts of the development process,
this could be troublesome in the future.
It now mounts to the same place as the uploads, results and web2py
database files

For some reason we are getting a hash sum mismatch on `apt-get update`
when building the image for the reporter.
This little hack helps prevent that error.

Some volumes were added to mount vidjil from the host.
Although this makes it easier to handle the contents of the vidjil
directory, it's not very user-friendly and also means the installation
is not self-sufficient.
Removing the volumes lets the containers default to the vidjil
installation already contained within the vidjil docker image.

The default database user for the docker installation was set to a
username we are not used to manipulating, and was a source of confusion.
Hopefully this change will lead to less confusion.

A missing semi-colon was causing an issue with a self-calling function.

Due to database pool size being hardcoded to 1, concurrent requests were
not possible and therefore slowed the app down considerably when several
users were trying to use the app at the same time. Often resulting in
timeouts.
Adding this parameter allows us to manage the pool size easily and
ensure users can post requests at the same time.

The cgis are needeed for some browser functionnalities.
They are compiled automatically into the vidjil package so we make the
most of this rather than recompile them for the browser package.

Changing the menu text to labels means we can delete some loops that
were designed to check the correct radio box onclick of the parent div.

Main server is now under the db.vidjil.org while app.vidjil.org
has just the frontend which can redirect anywhere.

Closes #2094

The text in the settings menu was included inside the inputs and
therefore clicking the text did not change the radio selection.
Changing the text into labels with the correct 'for' attribute fixes
this.