UNABLE TO READ 'playtime_seconds' error creates a infinite loop on a catalog add operation

[Verbocutane]

Description

A catalog containing an audio media with an invalid (or missing) "playtime_seconds" result in an incomplete catalog creation and an infinite error loop when trying to update the catalog (a new error log is created each 3 seconds until the administrator logs out).

To reproduce

Steps to reproduce the behavior:

1. Go to "Admin\Show Catalogs" section
2. On an existing catalog in which you have a media creating a "UNABLE TO READ 'playtime_seconds'" error, select action "Add" and click "GO"
3. You should see "Server-Sent events connection error. Re-connection.." pop-up coming up over and over again

Environment

• Ampache version: 6.2
• Web server + version: nginx/1.18.0 (Ubuntu)
• Server operating system: Ubuntu 22.04.3 LTS
• Client operating system: Windows 10 (Chrome 120.0.6099.130)

Client type

• Web

Logs

(local.catalog) -> Unable to get filesize for ...
(Ampache\Module\Util\VaInfo) -> UNABLE TO READ 'playtime_seconds'. This is probably a bad file ...

[lachlan-00]

can you do a compoer install / composer update?

i was getting this when getid3 wasn't installed correctly

[Verbocutane]

I have the latest composer version (V2.6.6) and the installation did not issue any error or warning

Note : it turns out that the files on which this error occurred were completely corrupted (empty files). After removing those files, the process of catalog creation finally finished.

Maybe a better behaviour would have been issue 1 error per corrupted file and skip it, but the error happened on the same file over and over again until it was removed (then it would loop on the next corrupted file until this one was also removed, and so on)

[lachlan-00]

empty file? i'll create an empty text file and try to replicate and bypass those issues

[lachlan-00]

db54ee6 should fix that now. a bad file will be added from what can be parsed but it also won't crash the process

[lachlan-00]

should be a return on the file size check too. that would stop it getting to the VaInfo class all together so that's set with fc296ec now too