esy-issues
esy-issues copied to clipboard
Easy ISSUES
@jordwalke writes: > Also, it turns out that OCAMLPATH must be semicolon separated on cygwin! I wonder if there's any other paths for ocaml/findlib that must be semicolon separated on...
`esy` uses `realpath` from `stdlib` when ejecting. (See [here](https://github.com/reasonml/esy/blob/esy/src/esy/buildEjectCommand/index.js#L400-L410)) According to @jordwalke it fails on `msys2`. IMO we should add an `#if` in our `realpath.c` and use [their implementation](https://sourceforge.net/u/earnie/msys2/msys2/ci/a949093a5e10123e445184b0a20e19e93e0b0658/tree/newlib/libc/sys/linux/realpath.c) if...
We should at least work well on winbash, even if we don't produce windows native executables yet.
We need to support `buildTimeDependencies` for that but there's a workaround suggested by @jordwalke: > So let's say I have two versions of Reason. And you're building the two separate...
opam package `conf-m4` checks if `m4` is available on a system. Instead we can package `m4` and use it instead. https://github.com/esy-ocaml/m4 The problem is that m4 build time is around...
I previously [commited](https://github.com/reasonml/esy/commit/0becd335c8a48a46945039d48c3815a7e9ec98be) a change that prevents irrelevant packages from populating any of the variables for install/bin/man etc. Currently, however, all of those packages still populate the `PATH`, which means...
The current [yarn fork](https://github.com/reasonml/esy) is nice because it has some built in resolving to fetch opam packages even though they don't have package.jsons. We can then use that project to...
- Constraint solver that generates `shrinkwrap.json`. - Constraint solver that generates yarn lockfile (stretch goal, maybe not needed).
Right now for local dependencies (symlinks or local file:// dependencies) `esy build` will rebuild them every time even if nothing changed, because we don't detect if anything changed. Instead, we...
It would be nice to get some clean, pure logic based tests that don't hit the network. @andreypopp Maybe your intermediate representation would be a convenient testing layer?