Re: hudsons hidden build errors

On Sun, 28 Nov 2010 09:51:46 -0700, Alex Rousskov
<rousskov_at_measurement-factory.com> wrote:
> On 11/27/2010 12:00 AM, Amos Jeffries wrote:
>> There are some build problems which are not being highlighted by the
>> hudson testing. Picking on FreeBSD-6.4 build because of problem #2.
> ...
>> libtool: compile: g++ -DHAVE_CONFIG_H -I../../.. -I../../../include
>> -I../../../lib -I../../../src -I../../include -I/usr/local/include
>> -I/usr/include -I/usr/include -I/usr/local/include -I/usr/include
>> -I/usr/include -Wall -Wpointer-arith -Wwrite-strings -Wcomments -Werror
>> -pipe -D_REENTRANT -D_FILE_OFFSET_BITS=64 -g -O2 -MT gadgets.lo -MD -MP
>> -MF .deps/gadgets.Tpo -c ../../../src/ssl/gadgets.cc -fPIC -DPIC -o
>> .libs/gadgets.o
>> ../../../src/ssl/gadgets.cc: In function `bool
>> addCnToRequest(Ssl::X509_REQ_Pointer&, const char*)':
>> ../../../src/ssl/gadgets.cc:17: warning: deprecated conversion from
>> string constant to `char*'
>> *** Error code 1
>> mv -f .deps/context_storage.Tpo .deps/context_storage.Plo
>> 1 error
>> *** Error code 1
>> 1 error
>> *** Error code 2
>> 1 error
>> *** Error code 1
>> 1 error
>> { test ! -d squid-3.HEAD-BZR || { find squid-3.HEAD-BZR -type d ! -perm
>> -200 -exec chmod u+w {} ';' && rm -fr squid-3.HEAD-BZR; }; }
>> ==================================================
>> squid-3.HEAD-BZR archives ready for distribution:
>> squid-3.HEAD-BZR.tar.gz
>> squid-3.HEAD-BZR.tar.bz2
>> ==================================================
>> buildtest.sh result is 0
>> BUILD: .././test-suite/buildtests/layer-02-maximus.opts
>> *** Error code 1
>> *** Error code 1
>> *** Error code 2
>> *** Error code 1
>> Build OK. Global result is 0.
>> "
>>
>> The execution of those test line appears to overwrite the $? magic
>> depended on by buildtest.sh.
>
> Agreed. We should not be building the distribution archives if the
> previous build targets fail.
>
> The "archives ready for distribution" lines appear to come from the
> automake-generated distcheck target. Is this an automake bug?

It came into view because a manual run of the testbed on west dies. Yet
the hudson run on same box indicates everything is fine.

I'm not sure where the bug is. Trying one re-run now after clearing the
workspace.

>
> I assume we bootstrap on FreeBSD. Is this bug specific to automake on
> FreeBSD? Can you compare the distcheck target of root Makefile.in on
> FreeBSD and Linux and see if there are any important differences? If
> yes, perhaps we can upgrade automake on the FreeBSD host?

This particular warning is about C-style cast being used. The other OS
don't complain about it so we can't see at present how they react.

Amos
Received on Mon Nov 29 2010 - 02:08:13 MST