Gabriel Scherer

The problem is that ocamlc and ocamlopt share some compilation units. For those, building ocamlopt will use the repo stdlib to build the `.cmx` file, while the `.cmi` file was...

After a short discussion with @Octachron and @damiendoligez, we are considering going the separate-build-dir route. I would like to avoid the horrible Makefile hacks in the style of `dynlink_compilerlibs`; one...

Is the code of your prototype pushed to a WIP-branch somewhere? My current understanding of the "shortest clean solution" would be to: - have separate build-artifact directories for boot-compiler artifacts...

One thing that I like with "multiple build directories" is that it would help with many other things, including: 1. cross-compilation (to avoid mixing object files written for the host...

Yes, the "relocatable configuration" is something we discussed in the past and, I also agree, would be good and solve other problems as well. (And it's simpler than hacking the...

Re. "relocatable configuration" (reading the compiler configuration from a file, instead of a configure-generated static module): this looks like something that @shindere might be interested in looking at.

Note: I propose a fix for the #2188-specific source of `.cmi` difference in #9592. (I believe this specific issue is blocking for the 4.11 release.)

Here is my take-away from the state of the discussion: - (1) I would like to be able to support situation where different compiler configurations produce different `.cmi` (in the...

I'm sorry to hear that. Maybe it would be possible to updatee the READMEs of https://github.com/clowdr-app/clowdr and https://github.com/clowdr-app/clowdr-app.github.io with a short banner, so that people asking themselves about the current...

After looking at this repo (from the ping I received on the other issue), I was meaning to post exactly this: I think you should give results for `Eval compute`...