You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
These features individually work, but if I enable both I get hundreds of build errors.
The repro I have is: clone rust repo, update library/test/Cargo.toml to add features = ['rustc-dep-of-std', 'extra_traits'] to libc. If you add only rustc-dep-of-std the crate builds fine. But if you add both you get a few hundred build errors, most in the form "X not found, try core::blah::X". Diff for the repro:
I was able to most of the build errors for my platform by just prefixing Debug with core::fmt:Debug, and the same for other symbols from core (Eq, PartialEq, ...). I also tried tweaking cfg_attrs in lib.rs but couldn't make this work with just cfg_attr tweaks.
Remaining issues are fixed by adding a few anonymous lifetime annotations, like &core::fmt::Formatter<'_> instead of &Formatter.
Why is this needed?
I was hoping to use the ctrlc crate in rustc's test runner, which uses nix, which uses libc with the 'extra_traits' feature. Without 'extra_traits' it's possible to use libc in rustc test runner without the 'rustc-dep-of-std' feature, but with 'extra_traits' we need 'rustc-dep-of-std' too.
I'm happy to submit a PR to add prefixes to symbols like Debug, Formatter, PartialEq, etc. but I'm not sure if that's the best way to fix this. It's easily possiblewith a combination of fd and sed, but the diff is large.
The text was updated successfully, but these errors were encountered:
We don't expect that these features are used together currently. The problem here is anonymous lifetime isn't available on old stable rustc (i.e. 1.25.0) so we should enable it only when rustc-dep-of-std and extra_traits. And these fixes make the code more complex, as you said, I'd like to avoid it generally :(
The patch is not complete: it works on my x86_64 Linux system, but modules for other platforms will also need to be updated.
If the diff looks OK and manageable I'll update the module for other platforms too and rely on CI to check my work as I only have access to x86_64 Linux.
These features individually work, but if I enable both I get hundreds of build errors.
The repro I have is: clone rust repo, update
library/test/Cargo.toml
to addfeatures = ['rustc-dep-of-std', 'extra_traits']
tolibc
. If you add onlyrustc-dep-of-std
the crate builds fine. But if you add both you get a few hundred build errors, most in the form "X not found, try core::blah::X". Diff for the repro:I was able to most of the build errors for my platform by just prefixing
Debug
withcore::fmt:Debug
, and the same for other symbols from core (Eq
,PartialEq
, ...). I also tried tweakingcfg_attr
s inlib.rs
but couldn't make this work with justcfg_attr
tweaks.Remaining issues are fixed by adding a few anonymous lifetime annotations, like
&core::fmt::Formatter<'_>
instead of&Formatter
.Why is this needed?
I was hoping to use the
ctrlc
crate in rustc's test runner, which usesnix
, which useslibc
with the 'extra_traits' feature. Without 'extra_traits' it's possible to use libc in rustc test runner without the 'rustc-dep-of-std' feature, but with 'extra_traits' we need 'rustc-dep-of-std' too.I'm happy to submit a PR to add prefixes to symbols like
Debug
,Formatter
,PartialEq
, etc. but I'm not sure if that's the best way to fix this. It's easily possiblewith a combination offd
andsed
, but the diff is large.The text was updated successfully, but these errors were encountered: