Re: [144336] trunk/dports/science/silo/Portfile

[David Strubbe]

Are you sure that is what this line meant? It is hard to believe that
ccache could be related to this error message of illegal syntax in a
Makefile. And, I just commented out the ccache line and the port seems to
work fine without that line, or with configure.ccache yes.

David

On Sat, Jan 9, 2016 at 2:31 AM, Ryan Schmidt 
wrote:

> On Jan 6, 2016, at 1:21 PM, dstru...@macports.org wrote:
> >
> > Revision
> > 144336
> > Author
> > dstru...@macports.org
> > Date
> > 2016-01-06 11:21:10 -0800 (Wed, 06 Jan 2016)
> > Log Message
> >
> > silo: Use compilers portgroup for Fortran variants. Clarify meaning in
> description. Update livecheck. Remove irrelevant comment. Add caution about
> MPI with HDF5.
>
> Well, the comment was relevant in that it documented the error message one
> got when ccache was used, thus explaining why it had been disabled for this
> port.
>
> > Modified Paths
> >
> >   • trunk/dports/science/silo/Portfile
> > Diff
> >
> > Modified: trunk/dports/science/silo/Portfile (144335 => 144336)
>
> > -# Makefile:152: *** missing separator.  Stop.
> > +compilers.choosefc f77 f90
> > +compilers.setup
> > +
> >  configure.ccacheno
>
>
___
macports-dev mailing list
macports-dev@lists.macosforge.org
https://lists.macosforge.org/mailman/listinfo/macports-dev

Re: [144336] trunk/dports/science/silo/Portfile

[Ryan Schmidt]

On Jan 9, 2016, at 9:45 PM, David Strubbe wrote:

>>> On Sat, Jan 9, 2016 at 2:31 AM, Ryan Schmidt wrote:
> 
>> On Jan 6, 2016, at 1:21 PM, dstru...@macports.org wrote:
>> >
>> > Revision
>> > 144336
>> > Author
>> > dstru...@macports.org
>> > Date
>> > 2016-01-06 11:21:10 -0800 (Wed, 06 Jan 2016)
>> > Log Message
>> >
>> > silo: Use compilers portgroup for Fortran variants. Clarify meaning in 
>> > description. Update livecheck. Remove irrelevant comment. Add caution 
>> > about MPI with HDF5.
>> 
>> Well, the comment was relevant in that it documented the error message one 
>> got when ccache was used, thus explaining why it had been disabled for this 
>> port.
>> 
>> > Modified Paths
>> >
>> >   • trunk/dports/science/silo/Portfile
>> > Diff
>> >
>> > Modified: trunk/dports/science/silo/Portfile (144335 => 144336)
>> 
>> > -# Makefile:152: *** missing separator.  Stop.
>> > +compilers.choosefc f77 f90
>> > +compilers.setup
>> > +
>> >  configure.ccacheno
> 

> Are you sure that is what this line meant? It is hard to believe that ccache 
> could be related to this error message of illegal syntax in a Makefile. And, 
> I just commented out the ccache line and the port seems to work fine without 
> that line, or with configure.ccache yes.

Yes I'm sure, in that I put it there when I initially added the port in 
https://trac.macports.org/changeset/78261 and I mentioned it in 
https://trac.macports.org/ticket/29325#comment:4.


I just now backdated the port to that revision, re-enabled ccache and changed 
"hdf5-18" to "hdf5", and verified that silo 4.8 did have this problem. Lines 
151-152 of the Makefile ended up as:

CC_FULLPATH = /opt/local/bin/ccache
/usr/bin/clang

and later in the same file:

CXX_FULLPATH = /opt/local/bin/ccache
/usr/bin/clang++

In other words, the MacPorts-provided CC and CXX values "ccache /usr/bin/clang" 
and "ccache /usr/bin/clang++" were erroneously expanded and corrupted.


Now, in silo 4.10.2, if I re-enable ccache, the corresponding lines of Makefile 
read:

CC_FULLPATH = /opt/local/bin/ccache

and:

CXX_FULLPATH = /opt/local/bin/ccache

This is still wrong -- these variables are evidently meant to be the absolute 
paths to the compilers, but they have instead been computed as the absolute 
path to ccache -- but at least they've fixed things to use only the first word. 
It looks like the only places these variables are used is to show a summary at 
the end of the configure phase:

Compiling Options:
 C Compiler /opt/local/bin/ccache
   CPPFLAGS -I/opt/local -I/opt/local/include
 CFLAGS -pipe -Os -arch x86_64 -D_LARGEFILE_SOURCE 
-D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wdeclaration-after-statement
LDFLAGS -L/opt/local/lib 
-Wl,-headerpad_max_install_names -arch x86_64
   LIBS -lhdf5  
Languages:
   C++ Compiler /opt/local/bin/ccache
   CXXFLAGS -pipe -Os -stdlib=libc++ -arch x86_64
   Fortran Compiler 
FCFLAGS -pipe -Os -m64
 FCLIBS 

And the same information gets written to the installed file 
/opt/local/lib/libsiloh5.settings as well.

It should probably be changed to directly use the unmodified values of CC and 
CXX and not attempt the error-prone process of trying to expand them into an 
absolute path, especially since it seems to serve no purpose other than to be 
informational to the user. Then ccache support for this portfile can be 
reenabled.


___
macports-dev mailing list
macports-dev@lists.macosforge.org
https://lists.macosforge.org/mailman/listinfo/macports-dev

Re: [144336] trunk/dports/science/silo/Portfile

[Ryan Schmidt]

On Jan 6, 2016, at 1:21 PM, dstru...@macports.org wrote:
> 
> Revision
> 144336
> Author
> dstru...@macports.org
> Date
> 2016-01-06 11:21:10 -0800 (Wed, 06 Jan 2016)
> Log Message
> 
> silo: Use compilers portgroup for Fortran variants. Clarify meaning in 
> description. Update livecheck. Remove irrelevant comment. Add caution about 
> MPI with HDF5.

Well, the comment was relevant in that it documented the error message one got 
when ccache was used, thus explaining why it had been disabled for this port.

> Modified Paths
> 
>   • trunk/dports/science/silo/Portfile
> Diff
> 
> Modified: trunk/dports/science/silo/Portfile (144335 => 144336)

> -# Makefile:152: *** missing separator.  Stop.
> +compilers.choosefc f77 f90
> +compilers.setup
> +
>  configure.ccacheno

___
macports-dev mailing list
macports-dev@lists.macosforge.org
https://lists.macosforge.org/mailman/listinfo/macports-dev