07:52 <Kaspar[m]> <DavidBrown[m]> Maybe someone here has some ideas. In one project I'm working on, the underlying OS does initialization through an array of structs that are linked in the same section. I can make this work by using #[link_section...], and #[used] to make sure the symbol ends up there. The problem is that it...

07:52 <Kaspar[m]> there is an unstable `#[used(linker)]` that solves this. alternatively, you can use an added linker script that marks the section `used`. without, linking is not guaranteed...

07:53 <JamesMunns[m]> <DavidBrown[m]> Maybe someone here has some ideas. In one project I'm working on, the underlying OS does initialization through an array of structs that are linked in the same section. I can make this... (full message at <https://catircservices.org/_irc/v1/media/download/AW-9mcdcZEcY9EidDkQpCDeOzsqLJExxcQmCUSUbYQ288evhYPZfN-PbAFfRERn8HLCxMOuD15Al7cLiYTC2eLC_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9hT0NSWVRCQlNpdWpBcWdjdlNyTVd4a2Y>)

08:13 <dirbaio[m]> <Kaspar[m]> there is an unstable `#[used(linker)]` that solves this. alternatively, you can use an added linker script that marks the section `used`. without, linking is not guaranteed...

08:13 <dirbaio[m]> https://github.com/rust-lang/rust/pull/140872

08:13 <dirbaio[m]> Are you using an old rustc? In 1.89 "used" is equivalent to "used(linker)"

08:14 <dirbaio[m]> [@d3zd3z:matrix.org](https://matrix.to/#/@d3zd3z:matrix.org)

09:08 <Kaspar[m]> <dirbaio[m]> Are you using an old rustc? In 1.89 "used" is equivalent to "used(linker)"...

09:08 <Kaspar[m]> wow, I missed that. probably allows us to us drop one of the linker scripts!

13:34 <DavidBrown[m]> So, the section is 'KEEP' in the linker script. And I'm using #[used] with 1.89. It does seem that with codegen-units = 1, as long as something brings the entire crate in (I call a C function in lib.rs from the C code. So, I think this will at least work. But, with the default codegen-units, the compiler breaks the crate in a large number of separate .o files inside the archive, and parts that aren't directly referenced

13:34 <DavidBrown[m]> get left out, don't have there init blocks. It, for example, means that something like a logger init that doesn't otherwise get called doesn't happen (things that are referenced from rust, like critical section, or global allocator do see to be brought in).

13:35 <DavidBrown[m]> ChatGPT helpfully recommended I just have a tree of init functions I call into each of these modules, which kind of defeats the point of using the underlying kernel's init mechanism.

14:18 <Kaspar[m]> <DavidBrown[m]> So, the section is 'KEEP' in the linker script. And I'm using #[used] with 1.89. It does seem that with codegen-units = 1, as long as something brings the entire crate in (I call a C function in lib.rs from the C code. So, I think this will at least work. But, with the default codegen-units...

14:18 <Kaspar[m]> If the crate is otherwise unused, does `use your_crate as _;` help?

14:20 <DavidBrown[m]> Kaspar[m]: Without codegen-units = 0, it doesn't help. With it, it is probably enough to bring the code in, not sure what symbol the use references. With linker gc on, it seems to pretty aggressively eliminate symbols.

14:21 <JamesMunns[m]> <DavidBrown[m]> So, the section is 'KEEP' in the linker script. And I'm using #[used] with 1.89. It does seem that with codegen-units = 1, as long as something brings the entire crate in (I call a C function in lib.rs from the C code. So, I think this will at least work. But, with the default codegen-units...

14:21 <JamesMunns[m]> Just checking, is your static pub and/or no_mangle?

14:22 <JamesMunns[m]> <DavidBrown[m]> So, the section is 'KEEP' in the linker script. And I'm using #[used] with 1.89. It does seem that with codegen-units = 1, as long as something brings the entire crate in (I call a C function in lib.rs from the C code. So, I think this will at least work. But, with the default codegen-units...

14:22 <JamesMunns[m]> If the static is "private" from Rust's perspective, and it can see that no one is using it, it is generally free to assume that it can be removed

15:51 <DavidBrown[m]> <JamesMunns[m]> If the static is "private" from Rust's perspective, and it can see that no one is using it, it is generally free to assume that it can be removed

15:51 <DavidBrown[m]> Isn't `#[used]` though supposed to keep the symbol. This just seems like an interesting consequence of the multiple codegen units.

16:00 <Kaspar[m]> <DavidBrown[m]> Isn't `#[used]` though supposed to keep the symbol. This just seems like an interesting consequence of the multiple codegen units.

16:00 <Kaspar[m]> If it is private, there might not be a symbol.

16:01 <DavidBrown[m]> Kaspar[m]: So, it does seem to show up in an object, and it is just a matter of making sure that the linker decides to bring that object in.

21:07 <thejpster[m]> <JamesMunns[m]> Just checking, is your static pub and/or no_mangle?

21:07 <thejpster[m]> Godbolt generally requires functions to be no_mangle in order for it to show you their assembly

21:11 <Mihael[m]> So i have a question i couldn't really find a solution for myself...... (full message at <https://catircservices.org/_irc/v1/media/download/AZomCTBh979LzEUfujq2-Lim2qsZM4CKUjD628TT6bJJNI_vTO3p4E20SBDqrXdRgv8zIrPPP3JFbOz7k5F-luK_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9ZTGxpVlBQclVvZUZ4b0VXUE9lS2hzb1k>)

21:14 <Mihael[m]> They're all small buffers but i can't simply reuse one bigger buffer due to the nature of data output

21:42 <i509vcb[m]> The mspm0 and imxrt-rs rooms both upgraded to 12, I think the new room addresses need to be setup for the space?

21:44 <firefrommoonligh> <Mihael[m]> So i have a question i couldn't really find a solution for myself......

21:44 <firefrommoonligh> Yes. I would do static buffers.

21:45 <firefrommoonligh> If you really need that... if you're not running out of memory, maybe defining them locally in the function as you're doing is fine?

22:15 <Mihael[m]> <firefrommoonligh> Yes. I would do static buffers.... (full message at <https://catircservices.org/_irc/v1/media/download/AVOeo-cDa7wtq8aGAeq5nuyMU5R4BFywXkRWsgBKNw8h-sAQqPCIvk13TfVpfsQEJknW4oKhkTHRTUi--53B1NG_8AAAAAAAAGNhdGlyY3NlcnZpY2VzLm9yZy9oaEhvSlV5YklRR1VrS2R1U2FFeEpGU1I>)

22:39 <adamgreig[m]> new blog post on heapless is live: https://blog.rust-embedded.org/heapless-091/ ( sgued )

23:19 <davidtwco[m]> <adamgreig[m]> probably best if we can gather thoughts from everyone here and/or next week and try to summarise them, and in particular please don't post on the RFC itself if you're not a wg member (at this time)

23:19 <davidtwco[m]> I don’t check Matrix too often but happy to answer small questions on the build-std RFC here too. It’s not necessarily written for the Rust user but rather the maintainers that need to implement it and understand its implications.