This affects the diagnostics and completions when editing javascript
code in the editor.

using pdfjs-3.11.174-legacy-dist.zip

Note: viewer.js is modified to do nothing in webViewerLoad (one line
with return "ERP5 patch: disable here" is added at the top of this
function) because we do the initialization in the gadget.
I think something similar was also done with the previous version.

The tests were extended to cover:

 - opening an empty or broken PDF: this should not crash
 - clicking save: this should not crash and the updated document should
   render properly. The save action replaces the ERP5 document content
   with the content from the editor, so if user for example added
   annotations or draw on the PDF, the modified version is saved.
   We don't use this feature and it's not so easy to test, so for now
   the test only covers the "it should not crash and it should not
   break the document".
 - entering wrong password for encrypted PDFs should not crash

This might fix flaky "testLogin". One theory is that a service worker
request to fill the cache happens while the selenium browser logs out,
which cause browser to keeps the authentication cookie.

This test waits for service worker to have filled the cache. We also
change the name (00) to run it first, so that subsequent tests are
executed with the cache already pre-filled.

This test is disabled and we don't use svn anymore

Introduce another assertImageMatchSnapshot which stops on failure,
following the behavior of other selenium assertions.

Also output more details when the dimensions are different.

See merge request nexedi/erp5!1838

See merge request nexedi/erp5!1833

This would catch potential problems only occuring when the location has
been granted.

Without this field, sync does not work when location was granted.

Fixes error in console:

Refused to load manifest from <URL> because it violates the following
Content Security Policy directive: "default-src 'none'".
Note that 'manifest-src' was not explicitly set, so 'default-src' is
used as a fallback.

Use the new KeyboardEvent instead of deprecated KeyEvent and
initKeyEvent, see
https://developer.mozilla.org/en-US/docs/Web/API/KeyboardEvent/initKeyEvent#specifications
and
nexedi/slapos!800 (comment 115488)

In old firefox dispatching a fake keydown event with keycode 13 (enter)
on an input field was enough to submit the corresponding form, like in
a real browser and like in webdriver, but in modern browsers this no
longer submits the form. Adjust the logic to submit the form to mimic
real browser behavior.
Co-authored-by: Romain Courteaud <romain@nexedi.com>

When I run this test on chrome, the div has two classes, spreadsheet
and jexcel_tab, this fixes a failure that the element is not found in
erp5_officejs_ui_test:testFunctionalOfficeJSWebTable

We were setting the value to a string with seconds like
`2010-01-01T00:00:00` and all the tests were testing that the value of
the input element was exactly the same (ie. with seconds), but [the spec]
mentions that value is:

 - A valid date string representing the date.
 - A U+0054 LATIN CAPITAL LETTER T character (T).
 - A valid time string representing the time, expressed as the shortest
  possible string for the given time (e.g. omitting the seconds
  component entirely if the given time is zero seconds past the minute).

The test works fine when running on firefox 68 (the one from testnodes),
because this version does not implements datetime-local and such field
behave like text inputs, but when running with a modern browser
implementing the spec, the test fails because the value from the fields
`.value` does not have seconds.

While the change to rjs_gadget_erp5_datetimefield_js is not strictly
necessary, because when setting the value on a datetime-local the value
will be normalized and having 00 seconds or omitting the seconds is same,
this is done to keep the test passing on old browsers without support
for datetime-local fields.

[the spec]: https://html.spec.whatwg.org/multipage/common-microsyntaxes.html#concept-datetime-local

so that we can see them on testnode.

Because they depend on the browser version, the window size, the
available fonts and maybe other things, snapshot tests basically only
work when running on testnode and fail when we run the tests from our
browsers locally. With this option, the difference of snapshot are not
considered a failure.

Also adjust snapshots, because this adds a checkbox in the upper frame,
the viewport of the test frame is now a little bit taller.

nexedi/erp5@3ea8d425 (comment 193284)

See merge request nexedi/erp5!1837

This fixes a regression from 3ea8d425 (core: don't show list of items
in initial business template install, 2023-10-10), this dialog was also
used for business template update and it was never showing the list of
item.

This code seems to have been pushed accidentally as part of 6e5e8e8c
(fixup! test: follow recent ImageMagick spec., 2023-10-11)

This reverts commit 14831fae.

As discussed in nexedi/erp5@14831fae (comment 188678)
this change breaks test and it seems more natural to implement this with
listbox's sort parameter.

This change will be pushed again, probably as part of
nexedi/erp5!1305

This is what browsers do when user selects a file.

testnode uses ProcessManager.killall to terminate all processes from a
path. To determine if a process is from a path, it looks at the command
line. This does not work for processes using setproctitle to change
their command line.

We can see in ps:

    $ ps -edf | grep nginx
    slapuse+ 115059  45574  0 16:14 ?        00:00:00 nginx: master process /srv/slapgrid/slappart46/t/cvt/i/0/tmp/shared/nginx/6d79cb0e7d81dce1be97eec8a5712f08/sbin/nginx -c /srv/slapgrid/slappart46/t/cvt/i/0/tmp/inst/T-0/etc/nginx-master-introspection.conf
    slapuse+ 115090 115059  0 16:14 ?        00:00:00 nginx: worker process

or by looking at cmdline, which is what psutil.Process.cmdline is using:

    $ cat /proc/115090/cmdline
    nginx: worker process

and that's why sometimes when cancelling a software release test while
it is running tests from a software using nginx, some processes are
leaked, they keep using the port and next test running on this testnode
fail.

In that case, killall is called with /srv/slapgrid/slappart46/t/cvt , we
can not find such process with cmdline, but we can extend this heuristic
to use the current working directory:

    $ ls -al /proc/115090/cwd
    lrwxrwxrwx 1 slapuser46 slapuser46 0 Oct 19 16:16 /proc/115090/cwd -> /srv/slapgrid/slappart46/t/cvt/i/0/tmp/inst/T-0

This also applies an optimization of only considering processes of the
current unix user.

Products.ERP5Types.patches.CMFCoreSkinnable: Only add portal_callables to skin selection when tool exists

Avoids a noisy log:
  WARNING Products.ERP5Type.patches.CMFCoreSkinnable Skin folder portal_callables is in selection list but does not exist.

As already documented in this code, unrestrictedTraverse provides a flat dict
as "request" argument. PAS plugins cannot work with such fake request, so
such _extractUserIds call will not succeed in authenticating the user, and instead
produces (suppressed) exceptions within PAS.
As a result, neither codepaths can be followed:
- PAS cannot find any user, hence "if len(user_list) > 0:" is false
- the "else" codepath starts with "request._auth", which obviously raises when
  request is a dict
So, reorder the code so that the nature of the request is checked before
either codepath is entered, skipping the bulk of this code and avoiding calling
into PAS.

The default soft limit of open file descriptors is usually set to 1024
in order to avoid breaking old software which still uses select. In many
projects we may need a higher limit: particularly in Wendelin based
projects we easily reach this limit. Before this patch it was therefore
necessary to either patch ERP5 in the project specific SR or to manually
increase the limit of the zope processes (or the parent supervisor) with
a tool like prlimit [1]. With this patch it becomes possible to increase the
soft limit to the hard limit with a command line argument of the zopewsgi bin.
This simplifies setting the soft limit for any Wendelin project.

[1] https://man7.org/linux/man-pages/man1/prlimit.1.html

/reviewed-by @vpelletier, @jerome
/reviewed-on nexedi/erp5!1827

test_suite and tests_require are for the deprecated `setup.py test`
command that we no longer use.

During initial install in most cases we just want to install everything,
showing the listbox to choose install options is not necessary. The
problem is that this listbox is really big when installing multiple
business templates.

This introduces a "detailed install" checkbox, that is unchecked by
default, to make it possible to show the list to choose from anyway.

so that we can view in inline with firefox for Zelenium tests

This was applied manually on http://www.erp5.org/dists/snapshot/test_bt5

See merge request nexedi/erp5!1830

See merge request nexedi/erp5!1826