Comment 15 for bug 1993819

Revision history for this message
Simon Chopin (schopin) wrote : Re: MIR: cargo, dh-cargo

First off, my bad, I didn't think of this MIR when I sponsored the upload, I should have.

From our perspective, it doesn't change much regarding the MIR, as the goal is to have the cargo binary package in main. Just because it's now built from src:rustc rather than src:cargo doesn't mean it magically was seeded, so the MIR is still needed.

AFAIK, the MIRs for the dependencies are still valid, the final cargo binary is pretty much the same as it was with the split source package (with minor differences that Zixing is in a better position to explain).

The source merge actually makes the toolchain more robust as it enforces using the same versions as the ones usually consumed together in the wild, reducing the possibilities of us encountering rare bugs due to version mismatch between cargo and rustc.

Now, that's from our perspective. However, there are two problems that I didn't see coming with the merge:

1/ It makes your life harder in terms of review, because now you need to figure out which part of the code is actually involved in cargo vs other, non-MIRed tools
2/ It blows up the vendored sources data, filling it with a much bigger set of dependencies that aren't relevant to the rustc binary (IOW the package that's *currently* in main)