Re: Reinplace
At 3:25 PM -0400 4/2/12, Craig Treleaven wrote:
At 5:14 AM +1000 4/3/12, Joshua Root wrote:
On 2012-4-3 04:01 , Jeremy Lavergne wrote:
Nope, none of those characters have special meanings to Tcl nor
to the regular expression engine. It looks like a perfectly valid
regular expression to me. Hopefully it will become clearer to me
when I actually try to build the port. Please hold on. My
computer is currently occupied upgrading another port. When it's
done, I'll give mythtv another go.
I'm wondering if ' versus " matters here. All the Tcl
documentation and all our existinting replace using "
Yes, that is going to be an issue. Single quotes don't mean anything
special to Tcl. Braces in Tcl do what single quotes do in sh.
Wait a minute. The line of code at issue is:
system "find ${configure.dir} -name \"Makefile\" -exec sed -i ''
's/ -L / /g' {} \\; "
If I replace the single quotes (') double-quotes ("), I would have
to escape them all. Is the whole string being passed to a shell for
expansion? Its not hard, so I'll try it that way...
That still doesn't explain, though, why find proceeds merrily for 22
of the 78 directories and then quits without error.
Yeah, exact same result with:
system "find ${configure.dir} -name \"Makefile\" -exec sed -i \"\"
\"s/ -L / /g\" {} \\; "
Except it is harder to read!
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Reinplace
At 5:14 AM +1000 4/3/12, Joshua Root wrote:
On 2012-4-3 04:01 , Jeremy Lavergne wrote:
Nope, none of those characters have special meanings to Tcl nor
to the regular expression engine. It looks like a perfectly valid
regular expression to me. Hopefully it will become clearer to me
when I actually try to build the port. Please hold on. My computer
is currently occupied upgrading another port. When it's done, I'll
give mythtv another go.
I'm wondering if ' versus " matters here. All the Tcl
documentation and all our existinting replace using "
Yes, that is going to be an issue. Single quotes don't mean anything
special to Tcl. Braces in Tcl do what single quotes do in sh.
Wait a minute. The line of code at issue is:
system "find ${configure.dir} -name \"Makefile\" -exec sed -i '' 's/
-L / /g' {} \\; "
If I replace the single quotes (') double-quotes ("), I would have to
escape them all. Is the whole string being passed to a shell for
expansion? Its not hard, so I'll try it that way...
That still doesn't explain, though, why find proceeds merrily for 22
of the 78 directories and then quits without error.
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Reinplace
At 12:56 PM -0500 4/2/12, Ryan Schmidt wrote: On Apr 2, 2012, at 12:39, Craig Treleaven wrote: Maybe that's the problem--it is not supposed to be a "regular expression"; it is a literal string of 'space-hyphen-capital ell-space'. Does hyphen have special meaning in a regex? Nope, none of those characters have special meanings to Tcl nor to the regular expression engine. It looks like a perfectly valid regular expression to me. Hopefully it will become clearer to me when I actually try to build the port. Please hold on. My computer is currently occupied upgrading another port. When it's done, I'll give mythtv another go. Ryan, for building, remember that you need qt-4mac +mysql +debug. Craig ___ macports-dev mailing list macports-dev@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Traversing the source directories
At 12:17 PM -0500 4/2/12, Ryan Schmidt wrote:
On Apr 1, 2012, at 20:17, Craig Treleaven wrote:
http://dl.dropbox.com/u/26787680/Portfile.2012Apr01
In this email I'll offer some suggestions on how the Portfile is
written. Later, I'll try to build the port, and in a second email
I'll try to help you figure out why there are empty "-L" arguments
and what's going wrong with your efforts to remove them.
github.setupMythTV mythtv v0.25-rc-0-g92f7d1f
The version number in the github.setup line should only be the
version number, typically as it appears in the tag name. There
should not be a "v" or other prefix, nor the git revision/branch
number. The github.setup procedure accepts a fourth argument for the
tag prefix, and the portgroup automatically handles the presence of
the randomish branch number in the folder name. So you want simply:
github.setupMythTV mythtv 0.25-rc v
OK, but when I went the the MythTV GitHub page and manually
downloaded the tarball:
https://nodeload.github.com/MythTV/mythtv/tarball/v0.25-rc
The exact file name I got was
MythTV-mythtv-v0.25-rc-0-g92f7d1f.tar.gz. When I plugged that in as
above, the port fetch worked.
configure.args-delete "-arch i386"
There is no "-arch i386" in configure.args, so this statement does nothing.
configure.cflags-delete "-arch i386"
configure.cflags-append "-m32"
Why delete "-arch i386" and append "-m32"? Aren't those the same thing?
configure.ldflags-append"-m32 -F/System/Library/Frameworks"
configure.cppflags-append "-m32"
configure.cxxflags-append "-m32"
configure.ldflags and configure.cxxflags similarly already have
"-arch i386" in them and thus shouldn't need "-m32".
configure.cppflags doesn't need arch flags because the preprocessor
doesn't compile anything.
Sorry, we went around this last week. I'm trying to recreate a
Perl-based build system (that bundles everything into an .app) and
there are notes that if you want a 32 bit build, put "-m32" into
those places. If I can get a working build, I'll experiment with
removing it.
configure.env-appendQMAKESPEC="macx-g++" \
QMAKE_LIBDIR_OPENGL="/opt/local/lib" \
Don't hardcode /opt/local; use ${prefix}.
Good catch.
configure.env-delete-arch i386 # messes up qmake?
That's not a valid environment specifier (which would always have
the form KEY=value). So I don't believe this statement does anything.
If you really want MacPorts not to add the -arch flags it normally
adds, clear the configure.cc_archflags configure.cxx_archflags
configure.ld_archflags variables. Normally you don't want to do that.
Thanks, I'll give that a go.
> system "find ${configure.dir} -name \"Makefile\" -print "
system "find ${configure.dir} -name \"Makefile\"
-exec sed -i '' 's/ -L / /g' {} \\; "
"find" can print and exec in the same invocation:
system "find ${configure.dir} -name \"Makefile\"
-print -exec sed -i '' 's/ -L / /g' {} \\; "
BTW, I wondered if OS X's metadata update and/or DropBox were to
blame. I've disabled both but no change.
The last three times, it has listed exactly 22 of the directories.
It appears to be going alphabetically down the root directory but I
see nothing special about the next directory in the list
(mythtv/filters).
Craig
--
--
Craig Treleaven, CA -- Clearview Consulting
(905) 829-2054 ctrelea...@cogeco.ca
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Reinplace
At 12:31 PM -0500 4/2/12, Ryan Schmidt wrote:
On Apr 1, 2012, at 08:53, Craig Treleaven wrote:
At 10:20 AM -0400 3/31/12, Jeremy Lavergne wrote:
> cd cd
"/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv"
sudo find . -name "Makefile" -exec sed -i '' 's/ -L / /g' {} \;
This is what MacPorts Reinplace extension is for, right? Would
this be the right syntax?
post-configure {
reinplace 's/ -L / /g' ${worksrcpath}/Makefile
}
Reinplace will do a recursive search from the specified
directory on down, right?
reinplace will only operate on the files you give to it. You need
to give it a complete list or you might use it in conjunction with
fs-traverse (see `man portfile`).
I'm struggling with this. Did I mention I'm new to MacPorts AND
tcl? ;) What I've got so far is this:
post-configure {
fs-traverse myfile ${configure.dir} {
if { [ file tail $myfile ] == "Makefile" } {
#ui_info "Removing any spurious -L
from ${myfile}"
reinplace "s/ -L / /g" ${myfile}
}
}
}
The trouble is that it does the reinplace on less than half the 20
or so files that need it.
Since you've subsequently noticed the same problem when running
"find", I suspect your regular expression is not correct. I will try
to build the port later and see if I can figure out what's going on.
Maybe that's the problem--it is not supposed to be a "regular
expression"; it is a literal string of 'space-hyphen-capital
ell-space'. Does hyphen have special meaning in a regex?
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Traversing the source directories
Hi:
I'm experiencing a weird error--neither fs_traverse or system "find .
blah" will consistently walk through _all_ of the directories in the
source of the port.
Background: for reasons unknown, I'm getting Makefiles that include
-L in the LIBS = line. There are 78 Makefiles and the
expedient thing to do is find them all and strip out the bad bits. I
ran the port configure step (this patching is done post-configure)
and it took MANY tries to finally get all 78 Makefiles searched and
patched. Many times, it was apparent that _some_ of the Makefiles
had been patched; but not all. I thought it was my inexperience with
tcl so I gave up on fs_traverse. Once or twice, it seemed "find .
-name "Makefile" -exec sed -i '' 's/ -L / /g' {} \;" _also_ stopped
midway but then it worked. I thought I'd found the magic formula.
Shortly after I achieved my first successful build of MythTV doing it
step by step. (Port fetch ... port install.) I wanted to start
fresh so I uninstalled and did a port clean then kicked off a port
install. The portfile and the debug log are linked below.
http://dl.dropbox.com/u/26787680/Portfile.2012Apr01
http://dl.dropbox.com/u/26787680/Post_configure%20error%202012Apr01.txt
The find command then quit after finding 22 of the 78 Makefiles.
(Because of the problems I've been having, I print them out to the
log first and then patch them all.) Of course, linking then fell
over after receiving one of the blasted -Ls that it hates so
much.
Is it possible that some of the Makefiles are still in use/open for
access when this command is running and that is causing fs_traverse
and/or find to silently give up the search? Or something else?
Should I do this step pre-build for some reason?
Any help gratefully accepted.
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Reinplace
At 10:20 AM -0400 3/31/12, Jeremy Lavergne wrote:
> cd cd
"/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv"
sudo find . -name "Makefile" -exec sed -i '' 's/ -L / /g' {} \;
This is what MacPorts Reinplace extension is for, right? Would
this be the right syntax?
post-configure {
reinplace 's/ -L / /g' ${worksrcpath}/Makefile
}
Reinplace will do a recursive search from the specified directory
on down, right?
reinplace will only operate on the files you give to it. You need to
give it a complete list or you might use it in conjunction with
fs-traverse (see `man portfile`).
I'm struggling with this. Did I mention I'm new to MacPorts AND tcl?
;) What I've got so far is this:
post-configure {
fs-traverse myfile ${configure.dir} {
if { [ file tail $myfile ] == "Makefile" } {
#ui_info "Removing any spurious -L
from ${myfile}"
reinplace "s/ -L / /g" ${myfile}
}
}
}
The trouble is that it does the reinplace on less than half the 20 or
so files that need it.
BTW, in debug output, it spews the following for every file that it fixes.
DEBUG: Executing proc-post-org.macports.configure-configure-0
DEBUG: euid/egid changed to: 0/0
DEBUG: chowned
/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv/Makefile
to macports
DEBUG: euid/egid changed to: 506/502
DEBUG: euid/egid changed to: 0/0
DEBUG: setting attributes on
/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv/Makefile
DEBUG: euid/egid changed to: 506/502
DEBUG: euid/egid changed to: 0/0
DEBUG: chowned
/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv/bindings/perl/Makefile
to macports
DEBUG: euid/egid changed to: 506/502
DEBUG: euid/egid changed to: 0/0
DEBUG: setting attributes on
/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv/bindings/perl/Makefile
DEBUG: euid/egid changed to: 506/502
I noticed the ".asroot" whatchamacallit but I don't seem to be able
to add that to post-configure. Is there some way to run the whole
segment with appropriate permissions?
Thanks for the hand-holding!
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Reinplace
Hi:
Sorry for all the newbie questions but could I'd appreciate some help.
Ten or so of the Makefiles include extraneous " -L " flags that I
need to strip out or linking fails. Manually, I did:
cd cd
"/opt/local/var/macports/build/_Users_craigtreleaven_MacPortsTemp_Myth.25/mythtv/work/mythtv-v0.25-rc-0-g92f7d1f/mythtv"
sudo find . -name "Makefile" -exec sed -i '' 's/ -L / /g' {} \;
This is what MacPorts Reinplace extension is for, right? Would this
be the right syntax?
post-configure {
reinplace 's/ -L / /g' ${worksrcpath}/Makefile
}
Reinplace will do a recursive search from the specified directory on
down, right?
Thanks,
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: DESTDIR not supported
At 4:09 PM +1100 3/31/12, Joshua Root wrote:
On 2012-3-31 09:54 , Ryan Schmidt wrote:
On Mar 30, 2012, at 15:45, Craig Treleaven wrote:
Well, it's been an interesting exercise getting this far and now
I find out that MythTV doesn't support DESTDIR. Is there a
standard approach to handling this problem? (Myth's configure
is >5,000 lines; I'm concerned about trying to path it.)
The MacPorts defailt is:
destroot.destdir DESTDIR=${destroot}
The standard is for build systems to support DESTDIR; if it
doesn't, all bets are off.
Most build systems still support a way to specify where to install
things, for example using a PREFIX or prefix variable. If so, you
can tell MacPorts something like:
destroot.destdir PREFIX=${destroot}${prefix}
If the build system's install target is too confused to do even
that properly, then you can throw it away and write your own
destroot block and do everything yourself, as Jeremy said. But I
would consider that a last resort.
You might also look at other distros that package MythTV; maybe one of
them fixed this.
It seems that INSTALL_ROOT= works but that's a good point. Do you
know of another port system that use destroot or similar? I just
looked and it seems FreeBSD does not.
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: DESTDIR not supported
At 5:54 PM -0500 3/30/12, Ryan Schmidt wrote:
On Mar 30, 2012, at 15:45, Craig Treleaven wrote:
Well, it's been an interesting exercise getting this far and now I
find out that MythTV doesn't support DESTDIR. Is there a standard
approach to handling this problem? (Myth's configure is >5,000
lines; I'm concerned about trying to path it.)
The MacPorts defailt is:
destroot.destdir DESTDIR=${destroot}
The standard is for build systems to support DESTDIR; if it doesn't,
all bets are off.
Most build systems still support a way to specify where to install
things, for example using a PREFIX or prefix variable. If so, you
can tell MacPorts something like:
destroot.destdir PREFIX=${destroot}${prefix}
If the build system's install target is too confused to do even that
properly, then you can throw it away and write your own destroot
block and do everything yourself, as Jeremy said. But I would
consider that a last resort.
Thanks, yes, PREFIX is supported. I'll give that a go.
Craig
___
macports-dev mailing list
macports-dev@lists.macosforge.org
http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
DESTDIR not supported
Hi: Well, it's been an interesting exercise getting this far and now I find out that MythTV doesn't support DESTDIR. Is there a standard approach to handling this problem? (Myth's configure is >5,000 lines; I'm concerned about trying to path it.) Craig ___ macports-dev mailing list macports-dev@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: Help with QT-based port?
At 8:47 AM -0500 3/29/12, Ryan Schmidt wrote:
On Mar 28, 2012, at 19:14, Craig Treleaven wrote:
I've run into a wall trying to port MythTV. The linker barfs complaining:
ld: -L must be immediately followed by a directory path (no space)
Debug-level log from my most recent attempt is here:
http://dl.dropbox.com/u/26787680/0.25%20macports%20build%20fail%202012Mar28.txt
I do agree with the error message... the log shows these arguments
being used on a g++ invocation:
-L -lQtGui_debug
Perhaps some directory name variable is not being set and is therefore empty.
Portfile in development* is here:
http://dl.dropbox.com/u/26787680/Portfile
fetch.type git
git.branch 92f7d1f51c
Does the 0.25-rc tarball posted on github not work? This is hundreds
of MB and takes a long time to fetch, which happens every time I
clean and try again, which is something I do often when
troubleshooting and/or developing a port, which makes it hard to
work on the port.
Sorry, I'm branch new to developing a port. Exactly, what should I
do to use the tarball? I agree that the git clone is quite slow.
license GPL
Can you be more specific about what version(s) of the GPL?
Need to research this. First, though, I want to get a working build! ;)
maintainers ctreleaven openmaintainer
Since you are not a committer, the maintainers line should contain
your actual (obfuscated) email address:
maintainers cogeco.ca:ctreleaven openmaintainer
Noted, thanks.
pre-configure { append worksrcpath "/mythtv" }
This is an odd way of setting worksrcpath, and fails if you ask
MacPorts to break the installation into separate steps, as in:
$ cd mythtv
$ sudo port configure
$ sudo port build
So you simply want:
worksrcdir ${name}-${version}/mythtv
Will correct.
You'll also want to remove the version line; the github.setup line
sets the version for you.
configure.env-appendMACOSX_DEPLOYMENT_TARGET="10.5" \
CFLAGS="-m32" \
CPPFLAGS="-m32" \
CXXFLAGS="-m32" \
LDFLAGS="-m32 -F/System/Library/Frameworks
-L/usr/lib -L${prefix}/lib" \
ECXXFLAGS="-m32"
I wouldn't think you'd need to specify -m32 everywhere, since
MacPorts already sets -arch i386 everywhere for you, which should be
the same thing.
I'm copying the existing build processes. Eventually, I want to
install some optional Perl and Python dependencies and I believe one
of them fails in a 64 bit environment. If I can get 32 bit working,
I'll go on and try 64 bit.
Note that MacPorts has special variables for most of these env vars,
which you should set or append to, instead of manually appending to
the environment:
macosx_deployment_target
configure.cflags
configure.cppflags
configure.cxxflags
configure.ldflags
Found those later, I'll change.
I'm surprised you need to specify -L/usr/lib, since /usr/lib is a
standard library directory that should be searched anyway. Or are
you just trying to force that /usr/lib is searched before
${prefix}/lib? If so, I'd be curious to know why, since that's the
opposite of what we usually want.
Ignorance on my part. I'll fix.
It is a problem that the port is building using "gcc" and "g++"; see:
https://trac.macports.org/wiki/UsingTheRightCompiler
Sorry, I've read the page but I don't understand the problem. My
output includes the following:
CXX='ccache /usr/bin/g++-4.2'
CC='ccache /usr/bin/gcc-4.2'
Is this a problem?
It appears to be compiling its own version of libavcodec. The
version that the ffmpeg port provides is not sufficient?
Correct. Myth compiles its own version of ffmpeg because there is
too much variability in arguments that various versions of ffmpeg
accept. For each major MythTV version, they resync ffmpeg.
See output of "port lint --nitpick" for some whitespace nits.
Will do.
Finally, I fail with:
dvdnav/dvdnav.c: In function 'dvdnav_open':
dvdnav/dvdnav.c:79: error: 'DVDNAV_SVN_REV' undeclared (first use in
this function)
dvdnav/dvdnav.c:79: error: (Each undeclared identifier is reported only once
dvdnav/dvdnav.c:79: error: for each function it appears in.)
Strange.
*Needs qt4-mac +mysql5. Takes 1 hour plus to build the
dependencies on my quad-core MBP.
Admittedly I have not done this yet. Guess I'd better get started.
In specifying the port dependency, how does one check that a
particular variant is present?
Also, some more searching suggests that qt ver 4.6.3 might work where
the default 4.7.4 doesn't. How does one get a list of the versions
of a port that may be installed?
Many thanks for
Help with QT-based port?
Hi: I've run into a wall trying to port MythTV. The linker barfs complaining: ld: -L must be immediately followed by a directory path (no space) Debug-level log from my most recent attempt is here: http://dl.dropbox.com/u/26787680/0.25%20macports%20build%20fail%202012Mar28.txt Portfile in development* is here: http://dl.dropbox.com/u/26787680/Portfile There are currently two building systems for Myth on Mac. First, a perl script has been in place for several years and is regularly maintained by the Myth devs. See oxs-packager.pl at: https://github.com/MythTV/packaging/tree/master/OSX/build Second, Myth also has a buildbot network set up including a Mac OS X slave. See: http://code.mythtv.org/buildbot/builders/master-osx-snow-leopard The buildbot shows quite clearly the environment variables and parameters passed to mythtv/config. Both systems build using the MythTV code straight from GitHub--no patching. I've spent a lot of time poring over that and the osx-packager.pl script but I can't come up with a combination of parameters that ends up with a happy linker. Myth uses Nokia's QT stuff so qmake is getting called in there. One difference is that both the buildbot and osx-packager.pl are using QT 4.6.3 whereas I've used MacPorts qt4-mac which is at 4.7.4. Could this be my problem? I've asked for help on the MythTV-dev list but the most active contributor to the Mac port is busy on other things and unwilling to spend any time on this. Any suggestions gratefully welcomed! Thanks, Craig *Needs qt4-mac +mysql5. Takes 1 hour plus to build the dependencies on my quad-core MBP. ___ macports-dev mailing list macports-dev@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
Re: MythTV port?
At 6:46 AM -0700 3/23/12, Bradley Giesbrecht wrote: On Mar 22, 2012, at 6:10 PM, Craig Treleaven wrote: > I'm going to *attempt* to be a maintainer--I'm very worried that I'm biting off far more than I can chew. If you are not familiar with MacPorts, start simple and build the port in phases. https://trac.macports.org/wiki/CommittersTipsAndTricks#TestingPortPhases http://guide.macports.org/#reference.phases You can work on anywhere convienent if you execute the port commands from within the directory. Create a directory somewhere named MythTV: $ mkdir ~/Desktop/MythTV $ cd ~/Desktop/MythTV $ touch Portfile Edit Portfile adding items 1 thru 13 here: http://guide.macports.org/#development.introduction Then test the fetch phase: $ sudo port -v fetch Continue on to patch etc... Thanks, I appreciate the pointers. I take it that it is not necessary to use SVN, at least to begin with? BTW, your last comment is about patching--if that is necessary, this is over! ;) But I know the code builds on OS X so I hope it is just a case of getting the right configure flags, etc. Craig ___ macports-dev mailing list macports-dev@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
