Re: nds broken by recent patches

[Jeff Law <jeffreyalaw@gmail.com>]

On 1/28/2022 10:13 PM, Mike Frysinger wrote:
> On 28 Jan 2022 09:36, Jeff Law wrote:
>> For example, someone mucked up the compiler for or1k-elf at the same
>> time you're doing your refactoring.  Now I've got a patch to fix the
>> compiler ICE, *but* I'm also getting a crap-ton of new testsuite
>> failures in the testsuite due to missing _init/_fini symbols.  I don't
>> know yet if that's a newlib issue or something else mucked up on the
>> compiler side.
> i had tried to run the newlib testsuite, but it didn't seem to work out
> of the box.  is there some trick/docs i'm missing to make them work ?
No idea.  I'm actually using the GCC testsuite.   The whole idea behind 
my tester is to catch GCC codegen issues closer to the point where 
they're introduced.  Of course for these embedded targets I need a C 
library & simulator.  Enter newlib and the old cygnus simulators....


So for a newlib target the sequence looks something like:

build & install binutils
build & install gcc & libgcc
build & install newlib
run gcc testsuite



>
>> The h8300-elf port in the last few days has started failing while
>> building newlib with:
>>
>> cp: cannot stat 'libc/crt0.o': No such file or directory
> i pushed a fix for this, so retry latest please
> https://sourceware.org/pipermail/newlib/2022/019024.html
Thanks.  I can confirm that's fixed.

>
>> The iq-2000-elf port has started failing building newlib in the last few
>> days with:
>>
>> make[3]: *** No rule to make target 'configure', needed by
>> 'config.status'.  Stop.
>> make[3]: Leaving directory
>> '/home/jlaw/jenkins/workspace/iq2000-elf/iq2000-elf-obj/newlib/iq2000-elf/libgloss/iq2000'
>> make[2]: *** [Makefile:132: stmp-bsp] Error 2
> i sent+pushed a fix for this
> https://sourceware.org/git/?p=newlib-cygwin.git;a=commit;h=580817ec0132265e6dfd0bb19b5deaf6b5866a35
Also confirmed as fixed.

>
>> And there's others.  Some are definitely on the GCC side (arc failures
>> for example) , but having so many things breaking at once is frustrating.
> if your CI builders are public, point me at them and i can flip through.
> i should be able to identify many of the libgloss/newlib ones that were
> my fault a bit quicker.
At this point we've covered those which are obviously due to the 
newlib/libgloss changes.  The current failures have to be examined with 
a debugger.  For example, bfin-elf seems to be failing after

> commit 754f8def0dfeeb43afa5a96ad1971fd0ef02c419
> Author: Mike Frysinger <vapier@gentoo.org>
> Date:   Sun Jan 23 01:10:33 2022 -0500
>
>     libgloss: merge stub arch configure scripts up a level
>
>     For about half the ports, we don't need a subdir configure script.
>     They're using the config/default.m[ht] rules, and they aren't doing
>     any unique configure tests, so they exist just to pass top-level
>     settings down to create the arch Makefile.  We can just as easily
>     do that from the top-level Mkaefile directly and skip configure.
>
>     Most of the remaining configure scripts could be migrated up to
>     the top-level too, but that would require care in each subdir.
>     So let's be lazy and put that off to another day.
>
But the failure mode is it looks like libgloss/bfin/syscalls.c got 
mis-compiled, but you have to dig into a failed GCC testsuite binary 
with gdb to make that determination.  On a positive note, you may have 
fixed this issue this morning with:

> commit 4b0e66093c0b48d5c363cf17c45ad9bf88ae526c
> Author: Mike Frysinger <vapier@gentoo.org>
> Date:   Fri Jan 28 05:12:54 2022 -0500
>
>     newlib: fix preprocessor checks


Anyway, if you want to take a peek at the CI/CD system

http://gcc.gnu.org/jenkins

Jeff