Temporary directory does not get removed

[braggpeaks]

weasyprint apparently creates a temporary directory but does not remove the directory afterwards.

Windows folder location: C:\Users\user.name\AppData\Local\Temp
Subfolder naming example: weasyprint-0qjsm58z

The folders are empty but I believe they should be removed to not leave unnecessary folders (I currently have 8000 of these weasyprint-* folders, don't ask 😂).

[liZe]

Hi!

The folders are supposed to be removed here, but:

• It is not guaranteed that __del__() methods are called for objects that still exist when the interpreter exits (quoting the Python documentation), and
• Windows likes to keep references to folders/files even when they’re not used anymore, preventing them to be deleted.

The "real" solution would be to only use data in memory (fonts are not that big). It should be possible to keep config files in memory thanks to FcConfigParseAndLoadFromMemory but there’s nothing to replace FcConfigAppFontAddFile.

If someone finds a better solution, I’ll be happy to try!

[seidnerj]

This is pretty annoying, it is cluttering the /tmp folder. This could also be the case if there was an errors when calling rmtree (as ignore_errors is set tot True).

[liZe]

This is pretty annoying, it is cluttering the /tmp folder.

We now use FcConfigParseAndLoadFromMemory, and I’ve just added a commit that avoids creating useless temporary folders. It means that you’ll only have the folders if you use @font-face rules, but the problem is not solved.

Unfortunately, there’s no function in Fontconfig to add fonts from memory (see almost 11-year old Fontconfig issue). Without such a function, as explained above, I didn’t find a reliable solution, ideas are welcome.

This could also be the case if there was an errors when calling rmtree (as ignore_errors is set tot True).

As far as I can remember, Windows was randomly trying hard to keep references to this folder, preventing rmtree from removing the folder.

Again, the real solution would be to avoid temporary folders, and it would require a new function in Fontconfig.

[seidnerj]

I understand. Well, if this will only happen when using @font-face then that solves this for me at least 🙂