Generate pdb file when strip=debuginfo

[ChrisDenton]

For msvc targets, treat strip=debuginfo the same as strip=none. See #122857 for why this is important. We should definitely not be stripping symbol names from external debug files in this case.

This is a much more minimal version of #115120. It only affects strip=debuginfo and nothing else.

[rustbot]

r? @estebank

rustbot has assigned @estebank.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[michaelwoerister]

Thanks for opening the PR, @ChrisDenton!

cc @rust-lang/wg-debugging

r? @michaelwoerister

[michaelwoerister]

@ChrisDenton, given that the immediate regression reported in #122857 is being addressed by the fix in Cargo, I'd prefer if we fixed the rustc side via #115120, which already has an MCP attached. It might take a bit longer to get #115120 merged, but I think even for this PR (#123031), we'd need to do an FCP and I'd like to avoid the organizational churn. I'll push for #115120 being resolved asap.

What do you think?

[ChrisDenton]

Sure. I'm ok with that. I do think the current behaviour of rustc itself is wrong but it's a long standing issue (cargo changes aside) so taking the time to do it right makes sense to me.

And personally I'll admit to always building with debug info so this isn't urgent for me.

[michaelwoerister]

Let's see if we can push #115120 over the finish line 🙂
#115120 (comment)