Issue 2551196
Created on 2022-02-17 22:38 by ngaba, last changed 2022-03-06 00:00 by rouilj.
File name |
Uploaded |
Description |
Edit |
Remove |
error.log.txt
|
ngaba,
2022-02-17 22:38
|
Error log |
|
|
msg7453 |
Author: [hidden] (ngaba) |
Date: 2022-02-17 22:38 |
|
When the history renderer of the web interface wants to print the change of a multilink property in the form
"propname: + added_item",
e.g.
"documents: + mydocument.docx",
it will crash, if the label property of the newly linked item is None. And in this case we get a cryptic error message in the browser (see the attached file error.log), which requires some time to debug.
I agree that no item should have empty labelprop, but users can do strange things... (In my scenario, the user made a filename empty.)
The problem is in the history() method of _HTMLItem class in cgi/templating.py (lines 1241-1248):
# if we have a label property, try to use it
# TODO: test for node existence even when
# there's no labelprop!
try:
if labelprop is not None and \
labelprop != 'id':
label = linkcl.get(linkid, labelprop)
label = html_escape(label)
Here html_escape(None) leads to an unhandled exception. I've added an
if label is None:
label = "[empty]"
block before calling html_escape() as a workaround, and it is fixed my issue. I note that setting a default_value for the multilinked Class in schema.py does not help either, because the above code ignores default_value...
|
msg7454 |
Author: [hidden] (rouilj) |
Date: 2022-02-21 03:33 |
|
Hi Nagy:
Thanks for identifying the issue and your patch. Can you try replacing your change with:
try:
if labelprop is not None and \
labelprop != 'id':
- label = linkcl.get(linkid, labelprop)
+ label = linkcl.get(linkid, labelprop,
+ default=self._("[la\
bel is missing]"))
label = html_escape(label)
and tell me if it still fixes the issue. Feel free to replace the string "[label is missing]".
Keep the self._() call as it allows the string to be localized. Given how get() is
implemented, I think this is the better way to fix the issue.
Also an added benefit is to allow the string to be translated or overridden using the
translation system.
I just wish I could figure out how to test this code. I have spent more than an hour
trying to figure out how to get a test and database entries right to trigger this code.
-- rouilj
|
msg7455 |
Author: [hidden] (ngaba) |
Date: 2022-03-05 21:10 |
|
Dear John,
Your patch also works for me, and it is more elegant.
|
msg7456 |
Author: [hidden] (rouilj) |
Date: 2022-03-06 00:00 |
|
Thanks for the confirmation Nagy.
I committed the fix for the multilink case you found. I also handled the link path
the same way. I also applied the fix to the plain() calls. I am not sure that it would
result in a crash, but it shouldn't be generating null anyway.
I still don't have a good way to test it, but...
Closing.
|
|
Date |
User |
Action |
Args |
2022-03-06 00:00:52 | rouilj | set | status: new -> fixed resolution: fixed messages:
+ msg7456 |
2022-03-05 21:10:17 | ngaba | set | messages:
+ msg7455 |
2022-02-21 03:33:56 | rouilj | set | keywords:
+ NeedsTesting assignee: rouilj messages:
+ msg7454 nosy:
+ rouilj priority: normal |
2022-02-17 22:38:13 | ngaba | create | |
|