Zulip Chat Archive

Stream: general

Topic: leanproject build deletes changes


Rémy Degenne (Mar 08 2021 at 08:49):

Doing leanproject build just deleted all my uncommited changes in the affected folder (including deleting a new file). I can reproduce it: I write something in a mathlib file, save, execute leanproject build, and the change is gone.
Is is a problem of my installation? Has leanproject been recently modified such that it is now supposed to do that? Any idea of how I can modify that behaviour or what additional info I could look for to understand what happens?

Rémy Degenne (Mar 08 2021 at 08:53):

a more detailed account of what I did: I created a branch l2_space on the mathlib github page, did leanproject get -b mathlib:l2_space, wrote some changes, did leanproject build, and my folder is back to the state it was in after leanproject get -b mathlib:l2_space.

Eric Wieser (Mar 08 2021 at 08:54):

https://github.com/leanprover-community/mathlib-tools/pull/96 fixes this. leanproject has always behaved like this, as far as I can tell

Rémy Degenne (Mar 08 2021 at 08:56):

It feels like it never did this to me before yesterday but I must be wrong.

Rémy Degenne (Mar 08 2021 at 08:57):

Thankfully I lost only some copy/pasted things from another folder, so I lost only 10min really.

Rémy Degenne (Mar 08 2021 at 09:00):

I'll be careful. Thanks for the reply!

Eric Wieser (Mar 08 2021 at 09:01):

I'm now just using that branch instead of the released version of mathlibtools, as I've blown away files multiple times this way and don't want to risk it any more.

Bolton Bailey (Apr 26 2021 at 18:03):

Yep, this just happened to me as well. Is it possible yet to update to a version of leanproject that doesn't have this problem?

Kevin Buzzard (Apr 26 2021 at 18:45):

Why not just do lean --make src? This has the extra advantage that you can add -M6000 and stop Lean from taking your desktop down.

Bolton Bailey (Apr 26 2021 at 18:50):

Great advice. If lean --make src is the better practice though, it seems like the guidance on the contribution guidelines page should say to do this rather than leanproject build. Can this be changed?

Rémy Degenne (Apr 26 2021 at 18:50):

Personally, I think the reason I did leanproject build and was caught by that issue once was that I followed the instructions on the "how to contribute page" https://leanprover-community.github.io/contribute/index.html . It states among other steps:

  • edit the necessary files, e.g. src/data/set/basic.lean
  • If you are anxious, leanproject build to check you didn't break anything.

Bolton Bailey (Apr 26 2021 at 18:51):

Wow, it seems we made the same comment at the same time

Kevin Buzzard (Apr 26 2021 at 19:00):

Oh I see! I hadn't realised that the claim was that the instructions are explicitly telling people to nuke their work!

Patrick Massot (Apr 26 2021 at 21:03):

Eric Wieser said:

https://github.com/leanprover-community/mathlib-tools/pull/96 fixes this. leanproject has always behaved like this, as far as I can tell

I'm pretty sure this bug was not present in the beginning. It was introduced when tweaking stuff. Eric, I think I was waiting for action from your side in this PR. Are you still interested in working on this?

Eric Wieser (Apr 26 2021 at 22:25):

Yeah, the ball is definitely in my court here. I realized that that PR messed up the workflow for people not working on mathlib, and haven't gotten around to getting back to it


Last updated: Dec 20 2023 at 11:08 UTC