Re: Compile/Link 101

2013-10-05 Thread Ryan Schmidt

On Oct 5, 2013, at 18:13, Craig Treleaven wrote:

> At 4:06 PM -0500 10/5/13, Ryan Schmidt wrote:
>> >>> My somewhat hazy recollection is that it might have started working in
> 4.4. Unfortunately I don't have that version handy to check.
>> 
>> http://trac.macports.org/ticket/36339#comment:3 says 4.4.1's clang doesn't 
>> support LIBRARY_PATH either. 4.5's does. So since the build system is so 
>> finicky you can blacklist macports-clang-2.9 macports-clang-3.0 and {clang < 
>> 421}
> 
> Thanks Ryan.  I'll have a go at that.
> 
> BTW, version information w.r.t. XCode 4.4 and 4.4.1 seems to be absent from 
> the wiki page.  Anyone have to details?
> 
> https://trac.macports.org/wiki/XcodeVersionInfo

Someone would have to download those versions of Xcode and their corresponding 
command line tools, install them, and get the requested info.

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


Re: Compile/Link 101

2013-10-05 Thread Craig Treleaven

At 4:06 PM -0500 10/5/13, Ryan Schmidt wrote:

 >>> My somewhat hazy recollection is that it might have started working in

 4.4. Unfortunately I don't have that version handy to check.


http://trac.macports.org/ticket/36339#comment:3 says 4.4.1's clang 
doesn't support LIBRARY_PATH either. 4.5's does. So since the build 
system is so finicky you can blacklist macports-clang-2.9 
macports-clang-3.0 and {clang < 421}


Thanks Ryan.  I'll have a go at that.

BTW, version information w.r.t. XCode 4.4 and 4.4.1 seems to be 
absent from the wiki page.  Anyone have to details?


https://trac.macports.org/wiki/XcodeVersionInfo

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


Re: Compile/Link 101

2013-10-05 Thread Ryan Schmidt

>>> My somewhat hazy recollection is that it might have started working in
>>> 4.4. Unfortunately I don't have that version handy to check.

http://trac.macports.org/ticket/36339#comment:3 says 4.4.1's clang doesn't 
support LIBRARY_PATH either. 4.5's does. So since the build system is so 
finicky you can blacklist macports-clang-2.9 macports-clang-3.0 and {clang < 
421}

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


Re: Compile/Link 101

2013-10-05 Thread Craig Treleaven
Sorry, I'd had a few different tabs open and must 
have pasted the link from the wrong one.


I've bent Michael's ear on this before, but 
Myth's build system is extremely hard for me to 
follow particularly due to QMake.  Different 
platforms get different qmkspec files--which are 
actually a chain of nested include files 
sometimes overriding each other.  I asked about 
this case because none of that was involved...and 
I still didn't understand why XCode 5.0 succeeded 
and XCode 4.3 failed.  At least now, I've got a 
clue.


OTOH, I checked the clang man page under XCode 5 
and it still doesn't mention the LIBRARY_PATH 
environment variable.


Craig

At 3:27 PM -0400 10/5/13, Michael Dickens wrote:

I think Craig is referring to the comment within the Portfile for
qt4-mac, about 2/5 of the way down under "Block various compilers".
MacPorts' build of Qt4 requires the use of CPATH and LIBRARY_PATH, and
post-configure I remove any -I${prefix}/lib entries to keep them from
appearing at inappropriate ordering during the build stage. So, I
blacklist all known compilers with this issue; it's the compiler that
makes the difference, much more than the version of XCode (though they
are related). And, during configuration Qt4 checks this property too.  I
think MythTV doesn't do path ordering carefully, but it's been a bit
since I looked at that project. - MLD

On Sat, Oct 5, 2013, at 03:09 PM, Joshua Root wrote:

 On 2013-10-6 00:53 , Craig Treleaven wrote:
 > At 11:03 PM -0400 10/4/13, Lawrence Velázquez wrote:
 >> MacPorts sets the LIBRARY_PATH environment variable in the configure
 >> phase, which sufficiently recent compilers take into account when
 >> linking. I don't know exactly when Clang started respecting it.
 >
 > Thank you!  From some searching, I found a bug report [1] for Qt4-mac
 > related to this problem. One comment suggested that XCode >= 4.5 and
 > clang >= 3.1 will respect LIBRARY_PATH.  Can someone confirm?
 >
 > [1] https://trac.macports.org/ticket/37878

 To which comment are you referring? AFAICT, all that ticket says
 regarding Xcode versions supporting LIBRARY_PATH is that it doesn't work
 with 4.3.3.

 My somewhat hazy recollection is that it might have started working in
 4.4. Unfortunately I don't have that version handy to check.



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


Re: Compile/Link 101

2013-10-05 Thread Michael Dickens
I think Craig is referring to the comment within the Portfile for
qt4-mac, about 2/5 of the way down under "Block various compilers".
MacPorts' build of Qt4 requires the use of CPATH and LIBRARY_PATH, and
post-configure I remove any -I${prefix}/lib entries to keep them from
appearing at inappropriate ordering during the build stage. So, I
blacklist all known compilers with this issue; it's the compiler that
makes the difference, much more than the version of XCode (though they
are related). And, during configuration Qt4 checks this property too. I
think MythTV doesn't do path ordering carefully, but it's been a bit
since I looked at that project. - MLD

On Sat, Oct 5, 2013, at 03:09 PM, Joshua Root wrote:

  On 2013-10-6 00:53 , Craig Treleaven wrote:

  At 11:03 PM -0400 10/4/13, Lawrence Velázquez wrote:

  MacPorts sets the LIBRARY_PATH environment variable in the configure
  phase, which sufficiently recent compilers take into account when
  linking. I don't know exactly when Clang started respecting it.

  Thank you!  From some searching, I found a bug report [1] for
  Qt4-mac
  related to this problem. One comment suggested that XCode >= 4.5 and
  clang >= 3.1 will respect LIBRARY_PATH.  Can someone confirm?
  [1] https://trac.macports.org/ticket/37878

  To which comment are you referring? AFAICT, all that ticket says
  regarding Xcode versions supporting LIBRARY_PATH is that it doesn't
  work
  with 4.3.3.
  My somewhat hazy recollection is that it might have started working
  in
  4.4. Unfortunately I don't have that version handy to check.
___
macports-dev mailing list
macports-dev@lists.macosforge.org
https://lists.macosforge.org/mailman/listinfo/macports-dev


Re: Compile/Link 101

2013-10-05 Thread Joshua Root
On 2013-10-6 00:53 , Craig Treleaven wrote:
> At 11:03 PM -0400 10/4/13, Lawrence Velázquez wrote:
>> MacPorts sets the LIBRARY_PATH environment variable in the configure
>> phase, which sufficiently recent compilers take into account when
>> linking. I don't know exactly when Clang started respecting it.
> 
> Thank you!  From some searching, I found a bug report [1] for Qt4-mac
> related to this problem. One comment suggested that XCode >= 4.5 and
> clang >= 3.1 will respect LIBRARY_PATH.  Can someone confirm?
> 
> [1] https://trac.macports.org/ticket/37878

To which comment are you referring? AFAICT, all that ticket says
regarding Xcode versions supporting LIBRARY_PATH is that it doesn't work
with 4.3.3.

My somewhat hazy recollection is that it might have started working in
4.4. Unfortunately I don't have that version handy to check.

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


Re: Compile/Link 101

2013-10-05 Thread Joshua Root
On 2013-10-6 00:53 , Craig Treleaven wrote:
> BTW, I realize I could set a -L flag in LDFLAGS but Myth includes
> modified versions of FFMPEG, dvdnav and a few other libraries.  The
> contents of LDFLAGS gets included early in the link line and therefore
> will try to link against any MacPorts-installed versions of these
> libraries. Not surprisingly, that never works out well.  ;)

This usually counts as a bug in the build system. The LDFLAGS from the
configure environment need to go after any -L flags for internal libs.
(Same deal for CPPFLAGS and -I flags for internal include dirs.)

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


Re: [111903] trunk/dports/kde

2013-10-05 Thread Ryan Schmidt

On Oct 5, 2013, at 09:13, ni...@macports.org wrote:

> Revision: 111903
>  https://trac.macports.org/changeset/111903
> Author:   ni...@macports.org
> Date: 2013-10-05 07:13:57 -0700 (Sat, 05 Oct 2013)
> Log Message:
> ---
> kdeedu ports: update to 4.11.2

> Modified: trunk/dports/kde/rocs/Portfile

> +#Clang of XCode 5 presently fails with a boost error 
> +#(fatal error: recursive template instantiation exceeded maximum depth of 
> 128)
> +#XCode 5.0.0 is 500.2.75
> +compiler.blacklist  {clang > 500}

We recently fixed this same error in vigra using:

configure.cxxflags-append -ftemplate-depth-1024

Maybe something similar will help rocs.

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


Re: Compile/Link 101

2013-10-05 Thread Craig Treleaven

At 11:03 PM -0400 10/4/13, Lawrence Velázquez wrote:

On Oct 4, 2013, at 9:39 PM, Craig Treleaven  wrote:

 I actually don't know how the test _can 
succeed_ with XCode 5?  There is no '-L' flag 
to indicate that the library is under 
/opt/local/lib.


MacPorts sets the LIBRARY_PATH environment 
variable in the configure phase, which 
sufficiently recent compilers take into account 
when linking. I don't know exactly when Clang 
started respecting it.


Thank you!  From some searching, I found a bug 
report [1] for Qt4-mac related to this problem. 
One comment suggested that XCode >= 4.5 and 
clang >= 3.1 will respect LIBRARY_PATH.  Can 
someone confirm?


[1] https://trac.macports.org/ticket/37878

BTW, I realize I could set a -L flag in LDFLAGS 
but Myth includes modified versions of FFMPEG, 
dvdnav and a few other libraries.  The contents 
of LDFLAGS gets included early in the link line 
and therefore will try to link against any 
MacPorts-installed versions of these libraries. 
Not surprisingly, that never works out well.  ;)


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


Re: New wiki page describing commit rules needed or nonsense?

2013-10-05 Thread Rainer Müller
On 2013-10-04 20:00, mk-macpo...@techno.ms wrote:
> I figured that a section about rules concerning properly formatted commit 
> messages are missing in the wiki (not to talk about the Guide), which is why 
> I created the following page:
> 
>   https://trac.macports.org/wiki/CommitRules
> 
> below the committers guide
> 
>   https://trac.macports.org/wiki/CommittersGuide
> 
> In case I am mistaken and the information is already present, but hidden 
> someplace else, please let me know.

We have point 8. there, "Commit messages should make clear what has 
been changed":

https://trac.macports.org/wiki/NewCommittersGuide#PortCommitterResponsibilities

I added a link to the new wiki page there as well. We could now 
probably remove the three points below that as they are covered in the 
new wiki page.

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