Revert "tools/testbuild.sh: Don't skip configure and distclean"

[Ouss4]

Summary

This reverts commit 05ff19d because now we are having a bunch of make: arm-nuttx-eabi-gcc: Command not found for some configs.

I've seen it first in #6197 but every PR has the same errors/messages.

Impact

Remove build error

Testing

CI will do the testing.

[Ouss4]

@xiaoxiang781216 the comment says:

Don't skip configure and distclean to improve the test coverage

I'm not sure how does this improve test coverage. Can you please explain it? Maybe we can implement it differently and not just revert your commit.

[xiaoxiang781216]

@xiaoxiang781216 the comment says:

Don't skip configure and distclean to improve the test coverage

I'm not sure how does this improve test coverage. Can you please explain it? Maybe we can implement it differently and not just revert your commit.

Check defconfig in the normalize form which is bypassed before my patch.

[xiaoxiang781216]

From the ci log, it look like that make try to invoke arm-nuttx-eabi-gcc in preconfig and distclean phase, but it shouldn't.

[Ouss4]

@xiaoxiang781216 the comment says:

Don't skip configure and distclean to improve the test coverage

I'm not sure how does this improve test coverage. Can you please explain it? Maybe we can implement it differently and not just revert your commit.

Check defconfig in the normalize form which is bypassed before my patch.

I see. But it looks like we can't get rid of those errors. The moment we call make, through sethost.sh when configuring, we get those. I tracked it to this: https://github.com/apache/incubator-nuttx/blob/fdef3a7b9220e53d898742251a6f5a721beb557b/arch/arm/src/armv7-m/Toolchain.defs#L259-L274

[xiaoxiang781216]

@xiaoxiang781216 the comment says:

Don't skip configure and distclean to improve the test coverage

I'm not sure how does this improve test coverage. Can you please explain it? Maybe we can implement it differently and not just revert your commit.

Check defconfig in the normalize form which is bypassed before my patch.

I see. But it looks like we can't get rid of those errors. The moment we call make, through sethost.sh when configuring, we get those. I tracked it to this:

https://github.com/apache/incubator-nuttx/blob/fdef3a7b9220e53d898742251a6f5a721beb557b/arch/arm/src/armv7-m/Toolchain.defs#L259-L274

@anchao could we delay the evaluation to the real used place?

[anchao]

@anchao could we delay the evaluation to the real used place?

@xiaoxiang781216 @Ouss4

I added clang's check in case of trying to print the libgcc path, this will get rid of this warning, please review this PR:
#6213

[Ouss4]

So the issue was the simply expanded variable at line 263: COMPILER_RT_LIB :=

@anchao why do we need that only for clang?

[xiaoxiang781216]

So the issue was the simply expanded variable at line 263: COMPILER_RT_LIB :=

@anchao why do we need that only for clang?

since only clang has this bug: ARM-software/LLVM-embedded-toolchain-for-Arm#109

[Ouss4]

So the issue was the simply expanded variable at line 263: COMPILER_RT_LIB :=
@anchao why do we need that only for clang?

since only clang has this bug: ARM-software/LLVM-embedded-toolchain-for-Arm#109

Thanks @xiaoxiang781216 @anchao. We don't need this one anymore, I'll close it.