cmake/common: Add language map entry for nasm

[res2k]

When determining the CMake compilers state, a CMakeLists.txt with these languages would be written:
project(CompInfo C NASM CXX)
Unfortunately, NASM isn't a CMake language. But ASM_NASM is - so map to that.

[res2k]

@mensinda, you have been assigned as a reviewer, so I hope you don't mind a question 😅

So first, the symptom this change wants to address is the warning "CMake Toolchain: Failed to determine CMake compilers state" being emitted, in certain circumstances where a subproject uses nasm sources.

Now I'd like to add a test that triggers the circumstances of the warning.
But what I'm struggling with: how to check for a warning being not emitted?
My first instinct would be using --fatal-meson-warnings, so the warning results in a failure; however, this only seems to be set for "warning" tests... and those generally seem to check for presence of some warning, not the absence.

So... do you happen to have any advice on how to approach this?

[dcbaker]

@res2k I don't know if mensinda will answer, they haven't been around for a while.

What you likely want to do is use a test.json, which has a field for looking for specific output from the meson setup command.

[res2k]

@dcbaker, thanks for chipping in.

I noticed you could check for the presence of some specific line in the output -
But I actually the opposite, check for absence.

Ie, fail case - the specific warning is emitted; success case - the specific warning is not emitted.

So I wonder if there's a way to do that with the current system ...

[dcbaker]

There is a way to search for a regex, I bet there's a way in python re to invert the condition?

[res2k]

There is a way to search for a regex, I bet there's a way in python re to invert the condition?

Actually, I noticed that there's apparently a way to check some output line occurs zero times, which seems to be what I want :)
Will give it a try.

[res2k]

@dcbaker, I managed to cobble together a test case that does what I want: produces "Failed to determine CMake compilers state" (and fails due to that) without the language map changes, but works warning-less with the latter change.
If you find some time to look at it, let me know what you think!

[dcbaker]

LGTM, thanks for adding this and the test

[dcbaker]

The error on the Ubuntu Rolling job looks relevant

[res2k]

The error on the Ubuntu Rolling job looks relevant

It's odd though. Pretty much looks like just nasm segfaulting, without really saying anything.

[dcbaker]

Ohhhh, that's awesome. Let me give that job a kick and see what happens.

[dcbaker]

Idea: that failing job is cross compiling to ARM, is there a chance that the ASM in question doesn't work an armv7?

[res2k]

Idea: that failing job is cross compiling to ARM, is there a chance that the ASM in question doesn't work an armv7?

Maybe. But TBH I would expect nasm to tell me that instead of crashing.
Also, the other cross-to-arm test (Ubuntu Rolling (--cross ubuntu-armhf.json --cross linux-mingw-w64-64bit.json, --unity=on, gcc, g++)) does not seem to have this problem. It's quite weird.

But I guess I'll disable the test for ARM anyway. Which is fine, I guess, the principle ("we don't write a bogus CMake language") is just as well demonstrated on other platforms.

[res2k]

But I guess I'll disable the test for ARM anyway. Which is fine, I guess, the principle ("we don't write a bogus CMake language") is just as well demonstrated on other platforms.

...then again, nasm is only "an asssembler for the x86 CPU architecture", so maybe the weird thing is that this cross-to-ARM sometimes worked?