templates: fix mismatched type size in simple

[colorglass]

Fix this #139. Using macro to select the build-in function in different type size.

For more background and discussion see also seL4/util_libs#176

[axel-h]

Based on seL4/util_libs#176 (comment) it seems
the minimal fix is simply using __builtin_clzl() then.

[colorglass]

Yes, we can just use CLZL() macro since it can handle the word type correctly.

[axel-h]

Looks good to me now.

You may want to raise an issue separately for supporting the LLP64 data model for 64-bit userland. Seem not many people have tries doing this using the "Windows" compiler.

[colorglass]

Sorry for my bothering.
This PR seems to be still linked to the closed seL4/util_libs#176, and the check got failed.
Is there any thing I can do to fix it? This is my first PR experience in github, and forgive my ignorance to this.

[lsf37]

The error in the build seems to be

/github/workspace/projects/util_libs/libutils/include/utils/builtin.h:30:36: error: expected ‘)’ before ‘sizeof’
     30 | compile_time_assert(long_is_32bits, sizeof(long) == 4)
        |                                    ^~~~~~~
        |                                    )

The syntax looks correct to me, but the compiler seems to disagree :-). Might need to track down the compile_time_assert definition that is used here.

Edit: just saw that is this is not even part of your change. Including utils/builtin.h already seem to trigger this, which is a bit unfortunate. Maybe there is some interaction with the inclusion of assert.h before. @axel-h do you know anything about that?

[colorglass]

/github/workspace/projects/util_libs/libutils/include/utils/builtin.h:30:36: error: expected ‘)’ before ‘sizeof’
     30 | compile_time_assert(long_is_32bits, sizeof(long) == 4)
        |                                    ^~~~~~~
        |                                    )

😢 Its my shame that I forgot to add the corresponded header file in that PR commit. And it was closed without the fix.

[axel-h]

The "Test with" was still in the PR description when this was pushed. Removed now and re-running, so CI will just use this PR and not the other one also.

[lsf37]

Good catch, thanks @axel-h. The link checker is unhappy with graphviz for some reason today, but that is unrelated. Good to merge from my side.

[lsf37]

Thanks for the contribution, @colorglass