cygport cross compilation

cygport cross compilation

[Ken Brown]

I've just started experimenting with using cygport for cross compiling, 
and I've come across two issues:

1. This is just a request: The latest cygport for Fedora appends i686 or 
x86_64 to the name of the working directory.  I would find it convenient 
if cygport did the same thing on Cygwin, at least if the --32 or --64 
option is specified.

2. I tried to do a build for x86-cygwin on x86_64-cygwin, but the 
compiler didn't work:

$ cat test.c
int
main ()
{
   return 0;
}

$ i686-pc-cygwin-gcc test.c
/tmp/ccaAaoj6.s: Assembler messages:
/tmp/ccaAaoj6.s:9: Error: invalid instruction suffix for `push'

Ken

Re: cygport cross compilation

[Yaakov (Cygwin/X)]

On 2014-03-19 11:32, Ken Brown wrote:
> I've just started experimenting with using cygport for cross compiling,
> and I've come across two issues:
>
> 1. This is just a request: The latest cygport for Fedora appends i686 or
> x86_64 to the name of the working directory.  I would find it convenient
> if cygport did the same thing on Cygwin, at least if the --32 or --64
> option is specified.

Actually, that is experimental code (based on a request from another 
cygport user) that was accidentally shipped when I had to reroll the 
source tarball for compatibility with F20/UnversionedDocDirs.  Would 
people like to see this done always, never, or only when cross-building?

> 2. I tried to do a build for x86-cygwin on x86_64-cygwin, but the
> compiler didn't work:
>
> $ cat test.c
> int
> main ()
> {
>    return 0;
> }
>
> $ i686-pc-cygwin-gcc test.c
> /tmp/ccaAaoj6.s: Assembler messages:
> /tmp/ccaAaoj6.s:9: Error: invalid instruction suffix for `push'

WFM.  Are you missing cygwin32-binutils?


Yaakov

Re: cygport cross compilation

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 960 bytes --]

On Mar 19 13:04, Yaakov (Cygwin/X) wrote:
> On 2014-03-19 11:32, Ken Brown wrote:
> >I've just started experimenting with using cygport for cross compiling,
> >and I've come across two issues:
> >
> >1. This is just a request: The latest cygport for Fedora appends i686 or
> >x86_64 to the name of the working directory.  I would find it convenient
> >if cygport did the same thing on Cygwin, at least if the --32 or --64
> >option is specified.
> 
> Actually, that is experimental code (based on a request from another
> cygport user) that was accidentally shipped when I had to reroll the
> source tarball for compatibility with F20/UnversionedDocDirs.  Would
> people like to see this done always, never, or only when
> cross-building?

Always wouldn't hurt, I think.


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: cygport cross compilation

[Ken Brown]

On 3/19/2014 2:28 PM, Corinna Vinschen wrote:
> On Mar 19 13:04, Yaakov (Cygwin/X) wrote:
>> On 2014-03-19 11:32, Ken Brown wrote:
>>> I've just started experimenting with using cygport for cross compiling,
>>> and I've come across two issues:
>>>
>>> 1. This is just a request: The latest cygport for Fedora appends i686 or
>>> x86_64 to the name of the working directory.  I would find it convenient
>>> if cygport did the same thing on Cygwin, at least if the --32 or --64
>>> option is specified.
>>
>> Actually, that is experimental code (based on a request from another
>> cygport user) that was accidentally shipped when I had to reroll the
>> source tarball for compatibility with F20/UnversionedDocDirs.  Would
>> people like to see this done always, never, or only when
>> cross-building?
>
> Always wouldn't hurt, I think.

That's my preference too.  If other people object, my second choice 
would be to have it happen whenever either --32 or --64 is specified 
(whether cross-building or not).

Ken

Re: cygport cross compilation

[Ken Brown]

On 3/19/2014 2:04 PM, Yaakov (Cygwin/X) wrote:
> On 2014-03-19 11:32, Ken Brown wrote:
>> 2. I tried to do a build for x86-cygwin on x86_64-cygwin, but the
>> compiler didn't work:
>>
>> $ cat test.c
>> int
>> main ()
>> {
>>    return 0;
>> }
>>
>> $ i686-pc-cygwin-gcc test.c
>> /tmp/ccaAaoj6.s: Assembler messages:
>> /tmp/ccaAaoj6.s:9: Error: invalid instruction suffix for `push'
>
> WFM.  Are you missing cygwin32-binutils?

It was installed, but "cygcheck -c" showed that it was incomplete. 
Reinstalling it fixed the problem.

Thanks.

Ken

[RFC] cygport: arch-specific workdir (was: cygport cross compilation)

[Yaakov (Cygwin/X)]

On 2014-03-19 13:04, Yaakov (Cygwin/X) wrote:
> On 2014-03-19 11:32, Ken Brown wrote:
>> I've just started experimenting with using cygport for cross compiling,
>> and I've come across two issues:
>>
>> 1. This is just a request: The latest cygport for Fedora appends i686 or
>> x86_64 to the name of the working directory.  I would find it convenient
>> if cygport did the same thing on Cygwin, at least if the --32 or --64
>> option is specified.
>
> Actually, that is experimental code (based on a request from another
> cygport user) that was accidentally shipped when I had to reroll the
> source tarball for compatibility with F20/UnversionedDocDirs.  Would
> people like to see this done always, never, or only when cross-building?

There hasn't been much comment on this.  Since it would be a visible 
change for package maintainers, I would appreciate more of a consensus. 
  Are there any objections to making the workdir always arch-specific 
(IOW name-ver-rel.arch)?


Yaakov

Re: [RFC] cygport: arch-specific workdir

[Achim Gratz]

Yaakov (Cygwin/X) writes:
>> Actually, that is experimental code (based on a request from another
>> cygport user) that was accidentally shipped when I had to reroll the
>> source tarball for compatibility with F20/UnversionedDocDirs.  Would
>> people like to see this done always, never, or only when cross-building?
>
> There hasn't been much comment on this.  Since it would be a visible
> change for package maintainers, I would appreciate more of a
> consensus. Are there any objections to making the workdir always
> arch-specific (IOW name-ver-rel.arch)?

I would rather have {build,inst,dist}/<arch> directories inside the
workdir so the same package could build for both x86 and x86_64 in one
workdir.


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Factory and User Sound Singles for Waldorf rackAttack:
http://Synth.Stromeko.net/Downloads.html#WaldorfSounds

Re: [RFC] cygport: arch-specific workdir

[Ken Brown]

On 3/27/2014 4:01 PM, Achim Gratz wrote:
> Yaakov (Cygwin/X) writes:
>>> Actually, that is experimental code (based on a request from another
>>> cygport user) that was accidentally shipped when I had to reroll the
>>> source tarball for compatibility with F20/UnversionedDocDirs.  Would
>>> people like to see this done always, never, or only when cross-building?
>>
>> There hasn't been much comment on this.  Since it would be a visible
>> change for package maintainers, I would appreciate more of a
>> consensus. Are there any objections to making the workdir always
>> arch-specific (IOW name-ver-rel.arch)?
>
> I would rather have {build,inst,dist}/<arch> directories inside the
> workdir so the same package could build for both x86 and x86_64 in one
> workdir.

That would be inconvenient for the situation where you have to do a new 
build (and bump the release number) for only one arch.  Also, there 
might be arch-specific patches, requiring arch-specific src and origsrc 
directories, so you don't really save any disk space by having a common 
workdir.

Ken

Re: [RFC] cygport: arch-specific workdir

[David Stacey]

On 27/03/14 18:50, Yaakov (Cygwin/X) wrote:
>> Actually, that is experimental code (based on a request from another
>> cygport user) that was accidentally shipped when I had to reroll the
>> source tarball for compatibility with F20/UnversionedDocDirs. Would
>> people like to see this done always, never, or only when cross-building?
>
> There hasn't been much comment on this.  Since it would be a visible 
> change for package maintainers, I would appreciate more of a 
> consensus.  Are there any objections to making the workdir always 
> arch-specific (IOW name-ver-rel.arch)?

I wrapped Cygport some time ago to give me a little extra functionality 
- for instance, for each package I maintain, it polls the project web 
page daily and does some regexp to see if a new version of the package 
had been released. If so, a new .cygport file is made and Cygport 
invoked to build, test and package the application.

So any change to Cygport would give me a little maintenance, but I could 
probably cope with a directory rename! Besides, some maintenance on my 
builder is a bit overdue: the tool is still uploading binaries to my 
Dropbox account and preparing [RFU] e-mails :-)

Dave.

Re: [RFC] cygport: arch-specific workdir

[David Rothenberger]

Yaakov (Cygwin/X) wrote:
> On 2014-03-19 13:04, Yaakov (Cygwin/X) wrote:
>> On 2014-03-19 11:32, Ken Brown wrote:
>>> I've just started experimenting with using cygport for cross compiling,
>>> and I've come across two issues:
>>>
>>> 1. This is just a request: The latest cygport for Fedora appends i686 or
>>> x86_64 to the name of the working directory.  I would find it convenient
>>> if cygport did the same thing on Cygwin, at least if the --32 or --64
>>> option is specified.
>>
>> Actually, that is experimental code (based on a request from another
>> cygport user) that was accidentally shipped when I had to reroll the
>> source tarball for compatibility with F20/UnversionedDocDirs.  Would
>> people like to see this done always, never, or only when cross-building?
> 
> There hasn't been much comment on this.  Since it would be a visible
> change for package maintainers, I would appreciate more of a consensus.
>  Are there any objections to making the workdir always arch-specific
> (IOW name-ver-rel.arch)?

I would favor making the workdir always arch-specific. It would allow me
to build my package for both archs at the same time. Currently, I have
to build for one, move the created packages, then rebuild for the other
arch.

-- 
David Rothenberger  ----  daveroth@acm.org

"Indecision is the basis of flexibility"
                -- button at a Science Fiction convention.

Re: [RFC] cygport: arch-specific workdir

[Yaakov (Cygwin/X)]

On 2014-03-27 13:50, Yaakov (Cygwin/X) wrote:
> On 2014-03-19 13:04, Yaakov (Cygwin/X) wrote:
>> On 2014-03-19 11:32, Ken Brown wrote:
>>> I've just started experimenting with using cygport for cross compiling,
>>> and I've come across two issues:
>>>
>>> 1. This is just a request: The latest cygport for Fedora appends i686 or
>>> x86_64 to the name of the working directory.  I would find it convenient
>>> if cygport did the same thing on Cygwin, at least if the --32 or --64
>>> option is specified.
>>
>> Actually, that is experimental code (based on a request from another
>> cygport user) that was accidentally shipped when I had to reroll the
>> source tarball for compatibility with F20/UnversionedDocDirs.  Would
>> people like to see this done always, never, or only when cross-building?

This is now in git master.


Yaakov

Re: [RFC] cygport: arch-specific workdir

[Achim Gratz]

Yaakov (Cygwin/X) writes:
> This is now in git master.

How about this change?  I'm usually having $arch set in my shell to
divert the build results to the proper release directories, and having
yet another naming convention doesn't really help with that.


bin/cygport.in: follow Cygwin convention for directory naming (x86 instead of i686)

diff --git a/bin/cygport.in b/bin/cygport.in
index 943d391..2ecfc25 100755
--- a/bin/cygport.in
+++ b/bin/cygport.in
@@ -672,7 +672,7 @@ unset restrict
 #
 ################################################################################
 
-declare -r workdir="${top}/${$PF}.${ARCH}";
+declare -r workdir="${top}/${$PF}.${ARCH/i686/x86}";
 declare -r srcdir="${workdir}/src";
 declare -r origsrcdir="${workdir}/origsrc";
 declare -r configdir="${workdir}/config";


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

SD adaptation for Waldorf microQ V2.22R2:
http://Synth.Stromeko.net/Downloads.html#WaldorfSDada