Cygport woes: CHOST not defined.

Cygport woes: CHOST not defined.

[Dave Korn]

    Hi list,

  Attempting to rebuild gcc4-4.3.4-3 from source, I got a bunch of error
messages during the install stage of cygport:

> Fixing libtool modules:
> /usr/lib/cygport/src_postinst.cygpart: line 763: test: i686-pc-cygwin: unary operator expected
    [ ... snip several more ... ]
> Stripping executables:
> /usr/lib/cygport/syntax.cygpart: line 191: test: i686-pc-cygwin: unary operator expected
> /usr/lib/cygport/syntax.cygpart: line 191: test: !=: unary operator expected
>         usr/bin/c++-4.exe
> /usr/lib/cygport/syntax.cygpart: line 191: test: i686-pc-cygwin: unary operator expected
> /usr/lib/cygport/syntax.cygpart: line 191: test: !=: unary operator expected
    [ ... snip several more ... ]

  Looking at the context, I infer that CHOST must be undefined:

> 760 # would otherwise be CHOST, but gcc ships $target libraries
> 761 case ${CTARGET} in
> 762 *-*-cygwin*|*-*-mingw*)
> 763     if test ${CTARGET} != ${CHOST} # cross-compiler, use sysroot
> 764     then
> 765         dodir /usr/${CTARGET}/sys-root/$(__target_prefix)/bin
> 766         mv ${ltlibdir}/${dlname} ${D}/usr/${CTARGET}/sys-root/$(__target_prefix)/bin/
> 767     else
> 768         origdlname=${dlname}

... and quite likely CBUILD as well:

> 190  __cross_compiling() {
> 191          return $(test ${CBUILD} != ${CHOST} || test ${CHOST} != ${CTARGET})
> 192  }

  This is basically a clean build - I renumbered the -3 extensions to -4, but
the cygport script is unchanged(*) so maybe there's something wrong in my
src_install override, to do with cygport's new cross-compiling features?

  I'll be grepping away trying to find out where these variables are supposed
to come from, but any hints that could speed up my debugging would be appreciated!

    cheers,
      DaveK
-- 
(*) -
http://www.mirrorservice.org/sites/sourceware.org/pub/cygwin/release/gcc4/gcc4-4.3.4-1-src.tar.bz2/gcc4-4.3.4-1.cygport?extract=true


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Cygport woes: CHOST not defined.

[Dave Korn]

On 17/03/2011 19:39, Dave Korn wrote:

>   Looking at the context, I infer that CHOST must be undefined:

  Ah, it's my cygport script.  For some reason I used to have a problem with
the setting of CHOST by the default cygport script, so it unsets it.
Hopefully I can just omit that with the newer cygport and everything will Just
Work.

    cheers,
      DaveK


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Cygport woes: CHOST not defined.

[Yaakov (Cygwin/X)]

On Thu, 2011-03-17 at 19:39 +0000, Dave Korn wrote:
>   Attempting to rebuild gcc4-4.3.4-3 from source, I got a bunch of error
> messages during the install stage of cygport:
> 
>     [ ... snip several more ... ]
> 
>   Looking at the context, I infer that CHOST must be undefined:
> ... and quite likely CBUILD as well:

Yes, CBUILD, CHOST, and CTARGET are used internally to support cygport's
various cross-compiling scenarios.  At some point I may make them
public, but either way they should be considered read-only.

BTW, do you need me to resend my patches for 4.5.2?


Yaakov



--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Cygport woes: CHOST not defined.

[Dave Korn]

On 18/03/2011 02:50, Yaakov (Cygwin/X) wrote:
> On Thu, 2011-03-17 at 19:39 +0000, Dave Korn wrote:
>>   Attempting to rebuild gcc4-4.3.4-3 from source, I got a bunch of error
>> messages during the install stage of cygport:
>>
>>     [ ... snip several more ... ]
>>
>>   Looking at the context, I infer that CHOST must be undefined:
>> ... and quite likely CBUILD as well:
> 
> Yes, CBUILD, CHOST, and CTARGET are used internally to support cygport's
> various cross-compiling scenarios.  At some point I may make them
> public, but either way they should be considered read-only.

  Indeed.  Can't remember what the long-ago bug was I ran into that made me
think unsetting it in the .cygport was the best solution, but the point is
moot anyway.

> BTW, do you need me to resend my patches for 4.5.2?

  Couldn't hurt, I know I had some stuff to dig through my inbox looking for
but if you wouldn't mind... <g>

  Once I've got linker plugin stuff sorted out upstream for
gcc-4.6/binutils-2.21.1, I'll build a 4.5.2 for the distro (and promote it to
curr, leaving 4.3.4-4 as prev) then start working on a 4.6.0 test package.

    cheers,
      DaveK




--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

gcc4 outstanding issues (was Re: Cygport woes: CHOST not defined.)

[Yaakov (Cygwin/X)]

On Fri, 2011-03-18 at 05:18 +0000, Dave Korn wrote:
> On 18/03/2011 02:50, Yaakov (Cygwin/X) wrote:
> > BTW, do you need me to resend my patches for 4.5.2?
> 
>   Couldn't hurt, I know I had some stuff to dig through my inbox looking for
> but if you wouldn't mind... <g>

My outstanding issues with gcc4:

1) /lib/cpp symlink for FHS compliance:
http://cygwin.com/ml/cygwin/2010-01/msg00601.html

2) fix-libtool-scripts-for-latest-gcc-runtimes.sh does nothing;
3) libgnat import libs are not installed;
4) Ada programs linked against shared libgnat do not terminate:
http://cygwin.com/ml/cygwin/2010-08/msg00412.html

5) Patch to fix Java NIO (*NOT* the patch above):
http://cygwin.com/ml/cygwin/2010-10/msg00562.html

6) Binaries should be linked against shared libintl:
http://cygwin-ports.git.sourceforge.net/git/gitweb.cgi?p=cygwin-ports/gcc4;a=blob_plain;f=config-rpath.patch

Except for (1) and (4), these are all fixed in Ports 4.5.2-1, and I've
been quite pleased with the results.

BTW, I wonder if we can begin mitigating (2) by not shipping the .la
files with GCC.  This would stop those paths from being hardcoded into
other .la files, so that eventually only the libtool scripts will need
to be fixed.

>   Once I've got linker plugin stuff sorted out upstream for
> gcc-4.6/binutils-2.21.1, I'll build a 4.5.2 for the distro (and promote it to
> curr, leaving 4.3.4-4 as prev) then start working on a 4.6.0 test package.

The sooner you can rebuild 4.5.2, the better; I've been using it for so
long now that I can't update some of my distro packages because they
depend on the newer GCC libraries.  Let me know if I can help in any
way.


Yaakov



--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: gcc4 outstanding issues (was Re: Cygport woes: CHOST not defined.)

[Charles Wilson]

On 3/18/2011 5:06 AM, Yaakov (Cygwin/X) wrote:
> BTW, I wonder if we can begin mitigating (2) by not shipping the .la
> files with GCC.  This would stop those paths from being hardcoded into
> other .la files, so that eventually only the libtool scripts will need
> to be fixed.

Yes, please.  I think this should be done, upstream (4.7?), as well.
There is no need for runtime libraries to have .la files IMO -- having
them needlessly complicates, and provides no benefit (is it even
supported, by gcc-devs, to dlopen/ltopen a runtime library, on all
platforms?)

--
Chuck

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple