Remember my request on the mailinglist regarding 'back' button? I've now found what makes users use the 'back' button:

Our tracker has several forms where instead of a 'menu' in html we use a 'field' with a (list) link to a help-popup, similar to what roundups own tracker does for the fields 'nosy list' and 'superseder'. The reason is that there are simply too many values to chose from.
Many of these fields produce a traceback when entering wrong values. In a tracker with 'debug' of the web-interface turned off, all the user sees is 'An Error has occurred'. In all the cases I have seen, an exception from the postgres backend

<class 'psycopg2.errors.InvalidTextRepresentation'>: invalid input syntax for type integer: "xxxx" ..
is seen (when debug is enabled in the web interface).

You can reproduce this in roundup's own tracker by entering "xxxxx" in the superseder field.

Hmm, I can't reproduce this with one of my trackers. It doesn't have a superseder field,
but it does have another issue multilink. In my tracker I see:

  "Error: you may only enter ID values for property seealso"

with a red background in the notification area.

The roundup tracker is running Roundup 1.6.1, so it's not a great test case but I do see
the problem.

What Roundup version are you running on your tracker? Also what template did you use to
create it?

On Mon, Dec 09, 2024 at 03:20:35PM +0000, John Rouillard wrote:
> 
> Hmm, I can't reproduce this with one of my trackers. It doesn't have a superseder field,
> but it does have another issue multilink. In my tracker I see:
> 
>   "Error: you may only enter ID values for property seealso"
> 
> with a red background in the notification area.

Do you have a postgres backend?
This may be a backend-specific error we do not catch.

> What Roundup version are you running on your tracker? Also what
> template did you use to create it?

This happens with current development. The latest rollout where this is
reproduceable include my performance-improvement merged branch.
Yes, these are my templates.

But please test our public roundup tracker and you will notice that it
occurs there, too.

Ralf
-- 
Dr. Ralf Schlatterbeck                  Tel:   +43/2243/26465-16
Open Source Consulting                  www:   www.runtux.com
Reichergasse 131, A-3411 Weidling       email: office@runtux.com

Hi Ralf:

> Do you have a postgres backend?
> This may be a backend-specific error we do not catch.

I do not. It's sqlite, which is more ... shall we say forgiving on typing issues.

Since you can reproduce with the current development, it's reasonable to think
its related to a postgres backend with psycopg2. IIRC there were some changes to the
exception structure in psycopg2.

InvalidTextRepresentation is a subtype of DataError:

 https://www.psycopg.org/docs/errors.html

Hi Ralf,

Have you had any luck tracking this down?

-- rouilj

On Sun, Jan 19, 2025 at 08:20:09PM +0000, John Rouillard wrote:
> 
> Hi Ralf,
> 
> Have you had any luck tracking this down?

Not yet, but it's quite high on my todo list.

Thanks
Ralf
-- 
Dr. Ralf Schlatterbeck                  Tel:   +43/2243/26465-16
Open Source Consulting                  www:   www.runtux.com
Reichergasse 131, A-3411 Weidling       email: office@runtux.com

Hi John, you write:

> Hmm, I can't reproduce this with one of my trackers. It doesn't have a superseder field,
> but it does have another issue multilink. In my tracker I see:
> 
>   "Error: you may only enter ID values for property seealso"

looking at the code this only happens when the class in question does not have a key (i.e. getkey() returns None). It specifically doesn't catch the case where the class *does* have a key like the superseder property linking to user (where the key is the username).

To reproduce the problem you may try to enter a non-existing username into the 'Superseder' property of *this* tracker.

I've currently only set up a test framework with postgres backend, I'll try to test this with another backend, too...

Hi Ralf:

I can reproduce it on one of my trackers using the watcher (nosy) field.
Which is a multilink to user. However I don't seem to get crash email.

But I can't on the other. Both use SQLite so:

> I've currently only set up a test framework with postgres backend, I'll
> try to test this with another backend, too...

It does look like there is an issue with at least one other RDBMS backend.

I can't explain why my I can't reproduce it by typing "fred" into the nosy(watcher)
field on my active development tracker. It is in sync with the current mercurial
tip 43899d99.

Thanks for tracking this down.