[OE-core] [PATCH] autotools.bbclass: output failed config.log

[Hongxu Jia]

On 5/15/19 9:42 PM, Adrian Bunk wrote:
> On Wed, May 15, 2019 at 08:30:03PM +0800, Hongxu Jia wrote:
>> On 5/15/19 7:08 PM, Adrian Bunk wrote:
>>> The end of config.log often contains expected error messages completely
>>> unrelated to the actual problem.
>> I am afraid you did not suffer pain from it, the most related message
>>
>> with do_configure failure is config.log. Take a example, as my previous
>>
>> commit [grub/grub-efi: fix unrecognized command line option
>>
>> '-pipe-Wno-error' in CFLAGS]
>>
>> It takes me hours to reproduce grub/grub-efi build faiure (only
>> ...
> I fully understand your pain, and have already suffered it many times.
>
>>> Debian autobuilders already dump config.log when when configuring failed,
>>> and there it is a common problem that people end up searching for the
>>> problem in the wrong places due to that.
>> It's another story, it is better that autobuilder provide sufficient and
>>
>> necessary info and data, not only config.log, but also local.conf,
>>
>> all available log.do_***, sources in ${B} and ${S}, even data in ${WORKDIR},
>>
>> result of `bitbake -e'
>>
>> but before we got there, the fix is best choice for do_configure failure
> There is no universal best choice here.
>
> Let me give you an example when building a more recent ofono locally
> with and without your patch.
>
> When a configure task fails, the last lines from log.do_configure
> are dumped to the shell.
>
> Without your patch I get when building in a shell:
> ...
> | checking whether to build shared libraries... yes
> | checking whether to build static libraries... no
> | checking for signalfd... yes
> | checking for dlopen in -ldl... yes
> | checking for glib-2.0 >= 2.32... yes
> | checking for dbus-1 >= 1.4... yes
> | checking for libudev >= 143... yes
> | checking for mobile-broadband-provider-info... yes
> | checking for ell >= 0.12... no
> | configure: error: Embedded Linux library >= 0.12 is required
> ...
>
> The problem is immediately visible.
>
> With your patch, one has to scroll up through the whole config.log for
> getting this information.
>
> And the config.log contains plenty of other (expected) errors a user
> might mistake as cause of the problem.
>
> For autobuilders dumping config.log can be OK if there is no better
> option available, but people looking at autobuilder logs should be
> aware what to scroll over when starting to look at the log.
>
> For anyone building manually in a shell these should really not be done,
> it has a high potential of confuding the user and the config.log is
> easily available.
>
> The best available option might be to have a knob with default off,
> and autobuilders might be configured to dump config.log.

After applying my fix, here is the sample

...

checking for i586-wrs-linux-gcc... i586-wrs-linux-gcc  -m32 -march=i586 
--sysroot=/buildarea1/hjia/wrlinux-1019/build_master-wr_qemux86_2019051509/build/tmp-glibc/work/i586-wrs-linux/grub/2.02-r0/recipe-sysroot
checking whether the C compiler works... no
configure: error: in 
`/buildarea1/hjia/wrlinux-1019/build_master-wr_qemux86_2019051509/build/tmp-glibc/work/i586-wrs-linux/grub/2.02-r0/build':
configure: error: C compiler cannot create executables
See `config.log' for more details
NOTE: The following config.log files may provide further information.
NOTE: 
/buildarea1/hjia/wrlinux-1019/build_master-wr_qemux86_2019051509/build/tmp-glibc/work/i586-wrs-linux/grub/2.02-r0/build/config.log
NOTE: This file contains any messages produced by compilers while
NOTE: running configure, to aid debugging if configure makes a mistake.

...

The content of config.log has prefix `NOTE:' and also tells user

See `config.log' for more dettails and balabala

//Hongxu

>> //Hongxu
> cu
> Adrian
>