DrRacket internal error for default save file location taken

[jasonhemann]

On Racket v9.0 [cs], on a MacOS 14.3 (23D56) machine, if it matters, I encountered the following when autosave attempted:

copy-file: destination exists
  source path: /Users/jhemann/Library/Preferences/PLT-autosave-toc.rktd
  destination path: /Users/jhemann/Library/Preferences/PLT-autosave-toc-save.rktd
  context...:
   /Applications/Racket v9.0/share/pkgs/gui-lib/framework/private/autosave.rkt:57:6: notify method in autosave-timer%
   /Applications/Racket v9.0/share/pkgs/gui-lib/mred/private/wx/common/timer.rkt:34:38
   /Applications/Racket v9.0/share/pkgs/gui-lib/mred/private/wx/common/queue.rkt:436:6
   /Applications/Racket v9.0/share/pkgs/gui-lib/mred/private/wx/common/queue.rkt:487:32
   /Applications/Racket v9.0/share/pkgs/gui-lib/mred/private/wx/common/queue.rkt:639:3

The proximate cause is, as it says, that copy-file throws an error when the PLT-autosave-toc-save.rktd file already exists. But I don't believe I've done anything that an ordinary user wouldn't've done. It seems like we were prevented from clobbering an autosave list, but it surprises me that the autosave feature is limited to one list, or that the lists aren't addable, or that we aren't forced to dispense with the existing autosave list before trying to autosave the next one. IIUC.

I tried to look around, but I couldn't find where the prefixed k:copy-file is coming from in racket/private/pre-base.rkt.

[rfindler]

Is it possible you have two DrRackets open?

In case you're curious, this is the relevant code.

[jasonhemann]

Something like that would be quite possible. I only have one drracket program installed

[jhemann:assignment_7547411_export] 12:30:01$ which -a drracket
/opt/homebrew/bin/drracket

But it's possible that I launched it twice, say once from my GUI launcher, and then once from my terminal. That could very well be the failure mode.

I got that far in the source. I tried then to track the copy-file method; I wanted to see where the error got thrown. Just to get my bearings. I could follow that to ./collects/racket/private/kw-file.rkt:174:10, but I got stuck trying to find k:copy-file used in 184. I can see that it must come from pre-base, because of the prefix (prefix-in k: "pre-base.rkt") but I have trouble finding that copy-file from within pre-base. It must be provided from somewhere else, but I couldn't (quickly) see where.

[rfindler]

I think the code at that level is just doing file-system things and probably the problem here is with managing-autosave things. Maybe I should be doing file-system locking to guard against multiple drrackets running at once.

[jasonhemann]

Your explanation of the code and diagnosis of the problem makes sense.

Maybe file-system locking to guard against multiple drrackets. Or it's also possible that multiple simultaneous drrackets is a useful feature, partially incompatible with autosaves, and that throwing up an error message like you're currently doing is already the Right Thing. If the latter case, the one thing I can think of that would improve over status quo is throwing some friendlier user-level error message in the case of multiple drrackets, describing collision. DrRacket internal error makes me think "file a bug" whereas user-level error signals "you goofed in a predictable, expected manner".

[rfindler]

Totally agree. No two ways about, this is a bug in DrRacket.