[PATCH v2] travis-ci: Disable C++ on Aarch64 container

[PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Philippe Mathieu-Daudé]

Travis-CI seems to have enforced memory limit on containers,
and the 'GCC check-tcg' job started to fail [*]:

  [2041/3679] Compiling C++ object libcommon.fa.p/disas_nanomips.cpp.o
  FAILED: libcommon.fa.p/disas_nanomips.cpp.o
  {standard input}: Assembler messages:
  {standard input}:577781: Warning: end of file not at end of a line; newline inserted
  {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
  {standard input}: Error: open CFI at the end of file; missing .cfi_endproc directive
  c++: fatal error: Killed signal terminated program cc1plus
  compilation terminated.

Until we have a replacement for this job on Gitlab-CI, disable
compilation of C++ files by forcing the c++ compiler to /bin/false
so Meson build system can not detect it:

  $ ../configure --cxx=/bin/false

  Compilation
                       C compiler: cc
                  Host C compiler: cc
                     C++ compiler: NO

[*] https://travis-ci.org/github/qemu/qemu/jobs/757819402#L3754

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
---
v2: Link to first line with error, describe Meson
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
---
 .travis.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.travis.yml b/.travis.yml
index 5f1dea873ec..b4b2d66fa4b 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -261,7 +261,7 @@ jobs:
           - genisoimage
       env:
         - TEST_CMD="make check check-tcg V=1"
-        - CONFIG="--disable-containers --target-list=${MAIN_SOFTMMU_TARGETS}"
+        - CONFIG="--disable-containers --target-list=${MAIN_SOFTMMU_TARGETS} --cxx=/bin/false"
         - UNRELIABLE=true
 
     - name: "[ppc64] GCC check-tcg"
-- 
2.26.2

Re: [PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Thomas Huth]

On 06/02/2021 21.05, Philippe Mathieu-Daudé wrote:
> Travis-CI seems to have enforced memory limit on containers,
> and the 'GCC check-tcg' job started to fail [*]:
> 
>    [2041/3679] Compiling C++ object libcommon.fa.p/disas_nanomips.cpp.o
>    FAILED: libcommon.fa.p/disas_nanomips.cpp.o
>    {standard input}: Assembler messages:
>    {standard input}:577781: Warning: end of file not at end of a line; newline inserted
>    {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
>    {standard input}: Error: open CFI at the end of file; missing .cfi_endproc directive
>    c++: fatal error: Killed signal terminated program cc1plus
>    compilation terminated.

If disabling C++ "fixes" the issue, ok ...
Otherwise, we should maybe rather limit the amount of parallel jobs there 
instead? (i.e. compiling with "make -j1" in the worst case?)

  Thomas

Re: [PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Philippe Mathieu-Daudé]

Hi Thomas,

On 2/8/21 6:54 AM, Thomas Huth wrote:
> On 06/02/2021 21.05, Philippe Mathieu-Daudé wrote:
>> Travis-CI seems to have enforced memory limit on containers,
>> and the 'GCC check-tcg' job started to fail [*]:
>>
>>    [2041/3679] Compiling C++ object libcommon.fa.p/disas_nanomips.cpp.o
>>    FAILED: libcommon.fa.p/disas_nanomips.cpp.o
>>    {standard input}: Assembler messages:
>>    {standard input}:577781: Warning: end of file not at end of a line;
>> newline inserted
>>    {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
>>    {standard input}: Error: open CFI at the end of file; missing
>> .cfi_endproc directive
>>    c++: fatal error: Killed signal terminated program cc1plus
>>    compilation terminated.
> 
> If disabling C++ "fixes" the issue, ok ...
> Otherwise, we should maybe rather limit the amount of parallel jobs
> there instead? (i.e. compiling with "make -j1" in the worst case?)

I exhausted my Travis-CI credits (frankly I don't plan to pay for it
with my own money). If the project expects developers to use Travis-CI,
we should think at some way to buy credits and give them to developers
on demand? [Cc'ing Stefan]

I tried to do my best to reproduce by limiting the amount of RAM, but
I don't know how to replicate Travis-CI hardware speed to match the 1h
time limit (the AArch64 hardware I have access to is certainly
different).

I'm happy with any patch (directly removing this job if it is not cared
of?) as long it silences the failure reports we get on IRC on each
merge.

Regards,

Phil.

Re: [PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Thomas Huth]

On 08/02/2021 10.16, Philippe Mathieu-Daudé wrote:
> Hi Thomas,
> 
> On 2/8/21 6:54 AM, Thomas Huth wrote:
>> On 06/02/2021 21.05, Philippe Mathieu-Daudé wrote:
>>> Travis-CI seems to have enforced memory limit on containers,
>>> and the 'GCC check-tcg' job started to fail [*]:
>>>
>>>     [2041/3679] Compiling C++ object libcommon.fa.p/disas_nanomips.cpp.o
>>>     FAILED: libcommon.fa.p/disas_nanomips.cpp.o
>>>     {standard input}: Assembler messages:
>>>     {standard input}:577781: Warning: end of file not at end of a line;
>>> newline inserted
>>>     {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
>>>     {standard input}: Error: open CFI at the end of file; missing
>>> .cfi_endproc directive
>>>     c++: fatal error: Killed signal terminated program cc1plus
>>>     compilation terminated.
>>
>> If disabling C++ "fixes" the issue, ok ...
>> Otherwise, we should maybe rather limit the amount of parallel jobs
>> there instead? (i.e. compiling with "make -j1" in the worst case?)
> 
> I exhausted my Travis-CI credits (frankly I don't plan to pay for it
> with my own money).

Yeah, same for me.

> I'm happy with any patch (directly removing this job if it is not cared
> of?) as long it silences the failure reports we get on IRC on each
> merge.

Ok, I wanted to send a PR with my other Travis->Gitlab patches today or 
tomorrow anyway, so I could also throw in your patch here to try to silence 
these failures (unless Alex wants to take this patch through his tree within 
the next days anyway)... then let's see whether disabling C++ is enough, and 
if not, we can still try my suggestion instead.

  Thomas

Re: [PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Philippe Mathieu-Daudé]

On 2/8/21 11:09 AM, Thomas Huth wrote:
> On 08/02/2021 10.16, Philippe Mathieu-Daudé wrote:
>> Hi Thomas,
>>
>> On 2/8/21 6:54 AM, Thomas Huth wrote:
>>> On 06/02/2021 21.05, Philippe Mathieu-Daudé wrote:
>>>> Travis-CI seems to have enforced memory limit on containers,
>>>> and the 'GCC check-tcg' job started to fail [*]:
>>>>
>>>>     [2041/3679] Compiling C++ object
>>>> libcommon.fa.p/disas_nanomips.cpp.o
>>>>     FAILED: libcommon.fa.p/disas_nanomips.cpp.o
>>>>     {standard input}: Assembler messages:
>>>>     {standard input}:577781: Warning: end of file not at end of a line;
>>>> newline inserted
>>>>     {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
>>>>     {standard input}: Error: open CFI at the end of file; missing
>>>> .cfi_endproc directive
>>>>     c++: fatal error: Killed signal terminated program cc1plus
>>>>     compilation terminated.
>>>
>>> If disabling C++ "fixes" the issue, ok ...
>>> Otherwise, we should maybe rather limit the amount of parallel jobs
>>> there instead? (i.e. compiling with "make -j1" in the worst case?)
>>
>> I exhausted my Travis-CI credits (frankly I don't plan to pay for it
>> with my own money).
> 
> Yeah, same for me.
> 
>> I'm happy with any patch (directly removing this job if it is not cared
>> of?) as long it silences the failure reports we get on IRC on each
>> merge.
> 
> Ok, I wanted to send a PR with my other Travis->Gitlab patches today or
> tomorrow anyway, so I could also throw in your patch here to try to
> silence these failures (unless Alex wants to take this patch through his
> tree within the next days anyway)... then let's see whether disabling
> C++ is enough, and if not, we can still try my suggestion instead.

FYI there is a v3 (20210207121239.2288530-1-f4bug@amsat.org) fixing
how AArch64 is written:
https://www.mail-archive.com/qemu-devel@nongnu.org/msg780124.html

OTOH we don't have any job explicitly disabling C++, right? (Not
sure we want one, just wondering).

Thanks,

Phil.

Re: [PATCH v2] travis-ci: Disable C++ on Aarch64 container

[Daniel P. Berrangé]

On Mon, Feb 08, 2021 at 10:16:19AM +0100, Philippe Mathieu-Daudé wrote:
> Hi Thomas,
> 
> On 2/8/21 6:54 AM, Thomas Huth wrote:
> > On 06/02/2021 21.05, Philippe Mathieu-Daudé wrote:
> >> Travis-CI seems to have enforced memory limit on containers,
> >> and the 'GCC check-tcg' job started to fail [*]:
> >>
> >>    [2041/3679] Compiling C++ object libcommon.fa.p/disas_nanomips.cpp.o
> >>    FAILED: libcommon.fa.p/disas_nanomips.cpp.o
> >>    {standard input}: Assembler messages:
> >>    {standard input}:577781: Warning: end of file not at end of a line;
> >> newline inserted
> >>    {standard input}:577882: Error: unknown pseudo-op: `.lvl35769'
> >>    {standard input}: Error: open CFI at the end of file; missing
> >> .cfi_endproc directive
> >>    c++: fatal error: Killed signal terminated program cc1plus
> >>    compilation terminated.
> > 
> > If disabling C++ "fixes" the issue, ok ...
> > Otherwise, we should maybe rather limit the amount of parallel jobs
> > there instead? (i.e. compiling with "make -j1" in the worst case?)
> 
> I exhausted my Travis-CI credits (frankly I don't plan to pay for it
> with my own money). If the project expects developers to use Travis-CI,
> we should think at some way to buy credits and give them to developers
> on demand? [Cc'ing Stefan]

We should just stop wasting our time here. Simply delete the travis
config right now and worry about adding more coverage to gitlab
separately.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|