Workflow question (changing codegen)

[Ben Gamari]

Ömer Sinan Ağacan <omeragacan at gmail.com> writes:

> My current workaround is this: I have a branch which is just master + the new
> file I've added. I first build it from a clean tree (git clean -xfd; then
> build), then switch to my branch, and run `make 1` in `compiler/`. That way I
> don't have to run ./configure (because the new file is already built and tracked
> by the build system) so the compiler version does not change and my stage 1
> compiler can use the libraries I built with master.
>
> I guess the root cause of this is that I have to run ./configure for the build
> system to track my new file, but doing that also updates the compiler version.
> Avoiding any of these (updating compiler version, or having to run configure
> when adding new files) would make this much easier.
>
For what it's worth, I sometimes just resort to manually editing the
files generated by ./configure to avoid having to reconfigure and
consequently rebuild.

Cheers,

- Ben

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 487 bytes
Desc: not available
URL: <http://mail.haskell.org/pipermail/ghc-devs/attachments/20190701/695443f1/attachment.sig>