Revert "arch: Replace ar and nm with gcc-ar and gcc-nm"

[yamt]

Summary

This reverts commit b05737d.

Because it broke clang-based builds.

Impact

Testing

[xiaoxiang781216]

Summary

This reverts commit b05737d.

Because it broke clang-based builds.

@yamt could you point out which config break? https://github.com/apache/incubator-nuttx/pull/4451/checks show that the change pass CI on macOS.

[yamt]

Summary

This reverts commit b05737d.
Because it broke clang-based builds.

@yamt could you point out which config break? https://github.com/apache/incubator-nuttx/pull/4451/checks show that the change pass CI on macOS.

clang build of sim/linux at least.
i don't think we are testing it on the ci.

[yamt]

i suppose LTO build will be an optional config.
i guess it makes more sense to make these conditional on, say, CONFIG_GCC_LTO.
how do you think?

[xiaoxiang781216]

i suppose LTO build will be an optional config.
i guess it makes more sense to make these conditional on, say, CONFIG_GCC_LTO.

Could be, but b05737d doesn't really enable lto, but make enable lto more easier.

how do you think?

For sim clang support, the better fix is SIM_TOOLCHAIN_CLANG/SIM_TOOLCHAIN_GCC and arch/sim/Toolchain.defs.

[yamt]

i suppose LTO build will be an optional config.
i guess it makes more sense to make these conditional on, say, CONFIG_GCC_LTO.

Could be, but b05737d doesn't really enable lto, but make enable lto more easier.

how do you think?

For sim clang support, the better fix is SIM_TOOLCHAIN_CLANG/SIM_TOOLCHAIN_GCC and arch/sim/Toolchain.defs.

i'm not sure it's better or not. but i agree it's consistent with the other boards.
are you going to fix this that way?

[yamt]

i suppose LTO build will be an optional config.
i guess it makes more sense to make these conditional on, say, CONFIG_GCC_LTO.

Could be, but b05737d doesn't really enable lto, but make enable lto more easier.

how do you think?

For sim clang support, the better fix is SIM_TOOLCHAIN_CLANG/SIM_TOOLCHAIN_GCC and arch/sim/Toolchain.defs.

i'm not sure it's better or not. but i agree it's consistent with the other boards.
are you going to fix this that way?

or are you asking me to do it? either way is ok for me.

[xiaoxiang781216]

We can add an issue. The change isn't huge if you don't have free time, I can do it in this week.

[yamt]

after thinking a bit, i guess the problem is not only for sim.
for example, it shouldn't be used in case of CONFIG_ARMV7M_TOOLCHAIN=CLANG, right?
i prefer this to be reverted for now unless we can fix it quickly.
(i don't think i can fix and test non-sim bits so quickly)

[xiaoxiang781216]

after thinking a bit, i guess the problem is not only for sim.
for example, it shouldn't be used in case of CONFIG_ARMV7M_TOOLCHAIN=CLANG, right?
i prefer this to be reverted for now unless we can fix it quickly.
(i don't think i can fix and test non-sim bits so quickly)

Fix by #4482

[xiaoxiang781216]

@yamt, since #4451 is related to #3836. I would propose to revert them in batch. Also, it's better to make a demo how to enable lto without gcc driver.

[yamt]

@yamt, since #4451 is related to #3836. I would propose to revert them in batch. Also, it's better to make a demo how to enable lto without gcc driver.

lto thing has been on my low-priority todo for a while. if anyone can beat me, it's welcome.

otoh, this one is more critical as it's breaking my (private) ci and workflows right now.

if you are going to merge #3836 without waiting for "lto demo", it's fine for me. (IMO it isn't reasonable to require me the demo in the first place.)
otherwise, let's fix this independently from #3836 .

[xiaoxiang781216]

@yamt, since #4451 is related to #3836. I would propose to revert them in batch. Also, it's better to make a demo how to enable lto without gcc driver.

lto thing has been on my low-priority todo for a while. if anyone can beat me, it's welcome.

otoh, this one is more critical as it's breaking my (private) ci and workflows right now.

if you are going to merge #3836 without waiting for "lto demo", it's fine for me. (IMO it isn't reasonable to require me the demo in the first place.)

It's fine without lto demo.

otherwise, let's fix this independently from #3836 .

Ok, let's merge this PR first.