Andy Polyakov

A bonus annoyance. [To be fair you're not exactly in position to catch it, because it's uncommon.] Here is output from attempt to compile CUDA project I maintain on Windows...

The major point is **usability** of the verbose output. Being a crate maintainer that relies on cc-rs I a) don't want users to face unrelated and confusing warnings, let alone...

The build script is programmed to not pass Wno-unused-function argument on Windows, so this shouldn't happen. Though it's assumed that user uses the msvc toolchain to compile. This is assumed...

So assertion is that with msvc as default toolchain pasta-msm build.rs script passes `-Wno-unused-function` to cl compiler. This shouldn't be possible and I can attest that it works for me....

As an **initial** remark. In respect to "potential memory leak." Execute ``` fn main(){} ``` under valgrind. To me it says "9 allocs, 8 frees, possibly lost: 0, still reachable:...

> In C it's a BSS symbol, and in assembly it's a common symbol: > > ``` > $ nm -og libblst.a | grep __blst_platform_cap > libblst.a:assembly.o:0000000000000004 C __blst_platform_cap >...

> it [not zeroing uninitialized values in .bss segment] was in violation of the language standard On a second though[t] I probably can't say this. The language standard stipulates that...

> The example that motivated this is an error when linking the static library into a Haskell program using `ghc`, the Haskell compiler. I can't reproduce the problem, not on...

> I'm wondering why you're using `.comm` in the assembly files instead of a simple external reference? Is it so that `assembly.o` can be used from the library without having...

> > The example that motivated this is an error when linking the static library into a Haskell program using `ghc`, the Haskell compiler. > > I can't reproduce the...