R SIG Fedora - Jun 2016 - problems compiling packages | 3.3.0 | Linux

On Tue, Jun 7, 2016 at 11:46 AM, Tom Callaway <tcallawa at redhat.com>
wrote:
> On 06/06/2016 05:04 PM, Brian Fallik wrote:
> > Apologies if this email is a top post. I'm not sure how to reply
to a
> > thread if I'm not subscribed.
> >
> > We recently encountered the same issue that Evan reported on 6/4 and I
> can
> > confirm that updated to 3.3.0-5 does address the build failure. Will
> > 3.3.0-5 make it into EPEL? If so, do you have any idea when? The
current
> > package is breaking our build process and I'm wondering if
it's worth the
> > effort to implement a workaround or not.
>
> Yes, this package is going through the update process right now. You can
> speed it along by logging into the Fedora Update system and giving
> positive karma:
>
> https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2016-93778a4eeb
>
> If the build gets +3 karma, it will go to stable immediately.
> Otherwise, it takes two weeks to be pushed to stable according to EPEL
> policies.
>
Thanks, Tom. I'll go try and help the karma process along.

brian

>
> ~tom
>
> => Red Hat
>
	[[alternative HTML version deleted]]

Hi Tom,

I'm seeing another similar build failure with the recent 3.3.0-5 and I'm
a
bit stumped about what's going on.

On a fresh VM, after installing 3.3.0-5 and then running:
  install.packages(c("lpSolveAPI), method="curl", repos="
https://cran.rstudio.com")
the build fails with:
  gcc: /builddir/build/BUILD/R-3.3.0/zlib-1.2.8/target/usr/lib64/libz.a: No
such file or directory
  gcc: /builddir/build/BUILD/R-3.3.0/bzip2-1.0.6/target/usr/lib64/libbz2.a:
No such file or directory
  gcc: /builddir/build/BUILD/R-3.3.0/xz-5.2.2/target/usr/lib64/liblzma.a:
No such file or directory
  gcc: /builddir/build/BUILD/R-3.3.0/pcre-8.38/target/usr/lib64/libpcre.a:
No such file or directory
  gcc:
/builddir/build/BUILD/R-3.3.0/curl-7.48.0/target/usr/lib64/libcurl.a: No
such file or directory

# rpm -qa | fgrep R
R-devel-3.3.0-5.el6.x86_64
R-3.3.0-5.el6.x86_64
R-core-3.3.0-5.el6.x86_64
R-java-devel-3.3.0-5.el6.x86_64
R-java-3.3.0-5.el6.x86_64
libRmath-devel-3.3.0-5.el6.x86_64
R-core-devel-3.3.0-5.el6.x86_64
libRmath-3.3.0-5.el6.x86_64

I'm confused how I'm seeing different behavior now form when I tested
3.3.0-5 yesterday, except in that case I'd originally started from 3.2.3
before upgrading to 3.3.0. Is it possible some build state leaked during
the upgrade from 3.2.3 to 3.3.0 allowing the build to succeed?

Another difference with 3.3.0-5 is that now only a single package is
failing, lpSolveAPI. Previously several packages all failed trying to link
against curl.

Any idea what's going on or suggestions for how to debug this further?

Thanks,
brian



On Tue, Jun 7, 2016 at 11:59 AM, Brian Fallik <brian at clypd.com> wrote:
>
>
> On Tue, Jun 7, 2016 at 11:46 AM, Tom Callaway <tcallawa at
redhat.com> wrote:
>
>> On 06/06/2016 05:04 PM, Brian Fallik wrote:
>> > Apologies if this email is a top post. I'm not sure how to
reply to a
>> > thread if I'm not subscribed.
>> >
>> > We recently encountered the same issue that Evan reported on 6/4
and I
>> can
>> > confirm that updated to 3.3.0-5 does address the build failure.
Will
>> > 3.3.0-5 make it into EPEL? If so, do you have any idea when? The
current
>> > package is breaking our build process and I'm wondering if
it's worth
>> the
>> > effort to implement a workaround or not.
>>
>> Yes, this package is going through the update process right now. You
can
>> speed it along by logging into the Fedora Update system and giving
>> positive karma:
>>
>> https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2016-93778a4eeb
>>
>> If the build gets +3 karma, it will go to stable immediately.
>> Otherwise, it takes two weeks to be pushed to stable according to EPEL
>> policies.
>>
>
> Thanks, Tom. I'll go try and help the karma process along.
>
> brian
>
>
>>
>> ~tom
>>
>> =>> Red Hat
>>
>
>
	[[alternative HTML version deleted]]

Brian Fallik <brian at ...> writes:
> 
> Hi Tom,
> 
> I'm seeing another similar build failure with the recent 3.3.0-5 and
I'm a
> bit stumped about what's going on.
> 
> On a fresh VM, after installing 3.3.0-5 and then running:
>   install.packages(c("lpSolveAPI), method="curl",
repos="
> https://cran.rstudio.com")
> the build fails with:
>   gcc: /builddir/build/BUILD/R-3.3.0/zlib-1.2.8/target/usr/lib64/libz.a: No
> such file or directory
>   gcc: /builddir/build/BUILD/R-3.3.0/bzip2-1.0.6/target/usr/lib64/libbz2.a:
> No such file or directory
>   gcc: /builddir/build/BUILD/R-3.3.0/xz-5.2.2/target/usr/lib64/liblzma.a:
> No such file or directory
>   gcc: /builddir/build/BUILD/R-3.3.0/pcre-8.38/target/usr/lib64/libpcre.a:
> No such file or directory
>   gcc:
> /builddir/build/BUILD/R-3.3.0/curl-7.48.0/target/usr/lib64/libcurl.a: No
> such file or directory


I also noticed this today (doing a large rebuild upgrading to bioconductor 3.3).

The culprit appears to be that FLIBS in /usr/lib64/R/etc/Makeconf (line 43)
still references these paths, and hence presumably it is only triggered if
you compile a package containing fortran code.

Deleting the paths from Makeconf makes the error go away (although I'm not
certain whether this means the correct libraries are being linked against -
this might be storing up problems for the future).

Gordon

(Centos 6.8, R 3.3.0-5)