global_asm expands arguments in comments

[gz]

I tried this code:

#![feature(global_asm)]
global_asm!{r#"
// ${num}
/* ${num} */
"#}

I expected to see this happen: Compiles without error (on nightly)

Instead, this happened (not sure if this is intended behavior or a bug):

error: there is no argument named `num`
 --> src/lib.rs:3:5
  |
3 | // ${num}
  |     ^^^^^

error: there is no argument named `num`
 --> src/lib.rs:4:5
  |
4 | /* ${num} */
  |     ^^^^^

[klensy]

Until raw string closed, all content interpreted as string, so that not a comments inside, but raw string. Or i misunderstand something.

[gz]

I ran into this because my global_asm!{} typically includes a .S file (with include_str!("file.S")), now that .S file is documented with comments and the comments also contained a ${num} string. This used to be fine with nightly-2021-03-25 or earlier but now throws this error after I upgraded to nightly-2021-06-01 -- so the behavior changed somewhere between these versions.

[klensy]

Perhaps #84107, try may-11 nightly.

[gz]

May 11th nightly works (does not produce the error)

rustc --version
rustc 1.45.0-nightly (9912925c2 2020-05-10)

[klensy]

That pr get merged May 14, 2021, 1:17 AM GMT+3, so you can try closest nightly.

rustc 1.45.0-nightly (9912925 2020-05-10)

2021-05-11, not 2020 )

[gz]

Ups -- sorry about that, I guess I'm still stuck in 2020!

works:

rustc --version
rustc 1.54.0-nightly (6d395a1c2 2021-05-13)

does not work:

rustc --version
rustc 1.54.0-nightly (1025db84a 2021-05-14)

[klensy]

$ git log 6d395a1..1025db8 --author=bors --oneline
1025db8 Auto merge of #85211 - Aaron1011:metadata-invalid-span, r=michaelwoerister
75da570 Auto merge of #83640 - bjorn3:shared_metadata_reader, r=nagisa
36a4d14 Auto merge of #85236 - nikic:update-llvm-submodule, r=cuviper
69b352e Auto merge of #85233 - FabianWolff:issue-85227, r=petrochenkov
91f2e2d Auto merge of #85190 - mati865:update-cc, r=Mark-Simulacrum
754d171 Auto merge of #85195 - Mark-Simulacrum:variant-by-idx, r=petrochenkov
17f30e5 Auto merge of #84107 - Amanieu:global_asm2, r=nagisa

So i guess i picked that one pr @Amanieu

[Amanieu]

#84107 changed global_asm! to accept arguments just like asm!. However this means that the string is now parsed the same way as a format string, which means {}, {1} and {name} are treated as placeholders for asm arguments.

The current workaround is to escape { and } using {{ and }} just like with format strings. One possibility would be to add options(raw) which would force treating the asm as a raw string instead of as a format string.

[gz]

Thanks the explanation. This isn't a problem for me in practice and looks like it's intentional after all, feel free to close this.

[fmease]

Closing as works-as-intended. However, I can reopen this issue if you'd like to pursue a potential options(raw) (alternatively you can open a new issue for that for the sake of clarity):

One possibility would be to add options(raw) which would force treating the asm as a raw string instead of as a format string.

[Amanieu]

options(raw) already exists.