Quarto should clean windows temp files

[kdheepak]

Bug description

I almost ran out of space today on my work computer and ended up looking at the temp folder and I see loads of folders created by quarto:

Here's a screenshot of me trying to delete them

Do you know why the operating system isn't cleaning this? I have quarto files from earlier this year in here.

Steps to reproduce

No response

Expected behavior

No response

Actual behavior

No response

Your environment

Windows 10 Enterprise

Quarto check output

No response

[cscheid]

We do clean up temporaries if the process finishes successfully (see eg src/command/render/render-services.ts).

One way in which temporary files might linger in the file system is if the process crashes. Assuming that's the source of these lingering files, (which are obviously a quarto bug!) we should probably have a single location in the file system that stores "files created by every session", which we could clean up at every quarto invocation.

[kdheepak]

I don't remember exactly how many folders there were (I didn't check the count at the bottom on Windows Explorer) and deleted them all now. But there were definitely in the several hundreds. I don't recall quarto crashing that many times (actually, I don't recall if I ever noticed quarto crashing). There were several instances over the last year where I'd Ctrl + C a session or where quarto would throw an error when it failed to render for some reason, but even that I don't know if it would amount to the number of files I saw here.

I tend to use quarto with computational cells a lot, with .qmd files with quarto preview. Is it possible that a new temp folder was being created for every time a notebook was rerendered in Julia?

[cscheid]

I think Ctrl + C on Windows specifically is to blame here.

Digging into this a little bit further:

• We have only one mention of quarto-session in our code;
• We have cleanup registration code for it;
• And that cleanup is called on both normal termination of the quarto code
• For abnormal termination of the quarto code, we also clean it up.. except on Windows. The commit that adds this change happened long ago, but the reason is "simply" that there are no signals on Windows, and so we don't do anything on abnormal termination 🙃

I think my earlier proposal, namely "we should probably have a single location in the file system that stores "files created by every session", which we could clean up at every quarto invocation" is really the solution.

[cscheid]

We could also attempt to do this cleanup in the Windows launcher, since it seems that Windows simply doesn't have great support for catching abnormal termination in-process.

[cscheid]

(@cderv, I assigned this to you because windows, just FYI)