Still struggling with Cygwin. One of the steps I still can't take is patch.

If I use the build-in patch from Cygwin emerge works fine. When I
emerge gentoo patch 2.6.1 the following emerges break while applying
patches.

I have some doubts, if anybody can help me here, but the build log is
a good example of recurring warnings in the different steps.

1.) The final warning of patch build:

* QA Notice: Package has poor programming practices which may compile
* fine but exhibit random runtime failures.
* src/pch.c:116: warning: implicit declaration of function ‘setmode’

2.) Short before end:

/home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
98: scanelf: command not found
/home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
186: scanelf: command not found

This prepstrip warning is one the is recurrent in different builds.
What does it mean? Can I ignore it?

3.) Somewhere above:

gcc: unrecognized option '-R/home/prefix/gentoo/usr/lib'
gcc: unrecognized option '-R/home/prefix/gentoo/lib'

That is the other recurring warning. What`s wrong? Something to ignore?

Al

Hi Al,

On 08/25/10 05:38, Al wrote:
> If I use the build-in patch from Cygwin emerge works fine. When I
> emerge gentoo patch 2.6.1 the following emerges break while applying
> patches.

It is unlikely one of those warnings is the reason for patch failing,
you'll have to debug the patch binary itself.

> 1.) The final warning of patch build:
>
> * QA Notice: Package has poor programming practices which may compile
> * fine but exhibit random runtime failures.
> * src/pch.c:116: warning: implicit declaration of function ‘setmode’

When you're in 32bit mode, it should be ok to ignore this one.

> 2.) Short before end:
>
> /home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
> 98: scanelf: command not found
> /home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
> 186: scanelf: command not found

Ignore it for now. This is specific to ELF systems, and in the end
there will be another platform-specific handling in prepstrip.

> 3.) Somewhere above:
>
> gcc: unrecognized option '-R/home/prefix/gentoo/usr/lib'
> gcc: unrecognized option '-R/home/prefix/gentoo/lib'

This will go away once you have bootstrapped.
Should be save to ignore too.

/haubi/
--
Michael Haubenwallner
Gentoo on a different level

2010/8/25 Michael Haubenwallner <haubi@gentoo.org>:
>
>> 2.) Short before end:
>>
>> /home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
>> 98: scanelf: command not found
>> /home/prefix/gentoo/usr/lib/portage/bin/ebuild-helpers/prepstrip: line
>> 186: scanelf: command not found
>
> Ignore it for now. This is specific to ELF systems, and in the end
> there will be another platform-specific handling in prepstrip.

Thank you very much Michael.

Yes, Cygwin is not ELF but COFF. I wounder if I have to change the
CHOST setting.

I tried "i686-pc-linux-gnu" so far, as Cygwin is a POSIX layer.

bootstrap-prefix.sh offers profiles for "i586-pc-interix" and
"i586-pc-winnt". Should I try one of theese?

Al

On 25-08-2010 12:25:15 +0200, Al wrote:
> > Ignore it for now. This is specific to ELF systems, and in the end
> > there will be another platform-specific handling in prepstrip.
>
> Thank you very much Michael.
>
> Yes, Cygwin is not ELF but COFF. I wounder if I have to change the
> CHOST setting.
>
> I tried "i686-pc-linux-gnu" so far, as Cygwin is a POSIX layer.
>
> bootstrap-prefix.sh offers profiles for "i586-pc-interix" and
> "i586-pc-winnt". Should I try one of theese?

Hmmm, I think you should use i686-pc-cygwin. That'll help make libtool
and the like do what they do best on Cygwin.

bootstrap-prefix.sh doesn't have support for this target (yet), since
there is no profile for it either. Though, I think it's going to harm
you more to set a wrong CHOST than hacking your way through it.

You'll have to create a profile for cygwin yourself (very limited, just
containing the CHOST should do, keyword = x86-interix or something is an
acceptable hack for now)


--
Fabian Groffen
Gentoo on a different level

>> bootstrap-prefix.sh offers profiles for "i586-pc-interix" and
>> "i586-pc-winnt". Should I try one of theese?
>
> Hmmm, I think you should use i686-pc-cygwin. *That'll help make libtool
> and the like do what they do best on Cygwin.
>

Thank you. I follow this advice.

Al

2010/8/25 Michael Haubenwallner <haubi@gentoo.org>:
> Hi Al,
>
> On 08/25/10 05:38, Al wrote:
>> If I use the build-in patch from Cygwin emerge works fine. When I
>> emerge gentoo patch 2.6.1 the following emerges break while applying
>> patches.
>
> It is unlikely one of those warnings is the reason for patch failing,
> you'll have to debug the patch binary itself.
>

I did, with a somewhat surprising result. This is the solution in short:

cp /bin/patch.exe.manifest /home/prefix/gentoo/tmp/usr/bin/
cp /bin/patch.exe.manifest /home/prefix/gentoo/usr/bin/

I set up a wiki page with a report of my investigations:

http://en.gentoo-wiki.com/wiki/Prefix/Cygwin

Al

"Al" <oss.elmar@googlemail.com> @ 2010-8-25 3:25 AM:
> Yes, Cygwin is not ELF but COFF.

I've almost always heard Windows executables called Portable Executables
(PEs) [1]. Though, according to the article, PEs are an evolution of COFFs.

Looking at a Cygwin binary (/usr/sbin/httpd2.exe from [2]), I do see the the
PE magic number [3], 0x4D 0x5A ("MZ"), not what appears to be the COFF magic
numbers [4], 0x4C 0x01.

[1] <http://en.wikipedia.org/wiki/Portable_Executable>
[2]
<http://cygwin.iphoneapps4free.org/release/apache2/apache2-2.2.6-1.tar.bz2>
[3] <http://en.wikipedia.org/wiki/Magic_number_(programming)>
[4] <http://www.delorie.com/djgpp/doc/coff/filhdr.html>

--
Christopher Warrington <chris@mygcw.net>

On 08/26/2010 08:42 AM, Christopher Warrington wrote:
> "Al" <oss.elmar@googlemail.com> @ 2010-8-25 3:25 AM:
>> Yes, Cygwin is not ELF but COFF.
>
> I've almost always heard Windows executables called Portable Executables
> (PEs) [1]. Though, according to the article, PEs are an evolution of COFFs.

Windows uses (that's the _real_ name ) PE/COFF. this means, PE
(personal executable) is the "container" (dll, exe), but the objects are
COFF (common object file format).

>
> Looking at a Cygwin binary (/usr/sbin/httpd2.exe from [2]), I do see the the
> PE magic number [3], 0x4D 0x5A ("MZ"), not what appears to be the COFF magic
> numbers [4], 0x4C 0x01.

yeah, right, but the objects that make up the binary all had the COFF
magic. the linker then puts those together and outputs a PE binary,
containing all the sections from each COFF object.

markus

>
> [1] <http://en.wikipedia.org/wiki/Portable_Executable>
> [2]
> <http://cygwin.iphoneapps4free.org/release/apache2/apache2-2.2.6-1.tar.bz2>
> [3] <http://en.wikipedia.org/wiki/Magic_number_(programming)>
> [4] <http://www.delorie.com/djgpp/doc/coff/filhdr.html>
>

On 08/26/2010 10:54 AM, Markus Duft wrote:
> On 08/26/2010 08:42 AM, Christopher Warrington wrote:
>> "Al" <oss.elmar@googlemail.com> @ 2010-8-25 3:25 AM:
>>> Yes, Cygwin is not ELF but COFF.
>>
>> I've almost always heard Windows executables called Portable Executables
>> (PEs) [1]. Though, according to the article, PEs are an evolution of COFFs.
>
> Windows uses (that's the _real_ name ) PE/COFF. this means, PE
> (personal executable) is the "container" (dll, exe), but the objects are
> COFF (common object file format).

hah - as haubi pointed out correctly, it's Portable Executable, of
course, not Personal Executable - fail :]

markus

>
>>
>> Looking at a Cygwin binary (/usr/sbin/httpd2.exe from [2]), I do see the the
>> PE magic number [3], 0x4D 0x5A ("MZ"), not what appears to be the COFF magic
>> numbers [4], 0x4C 0x01.
>
> yeah, right, but the objects that make up the binary all had the COFF
> magic. the linker then puts those together and outputs a PE binary,
> containing all the sections from each COFF object.
>
> markus
>
>>
>> [1] <http://en.wikipedia.org/wiki/Portable_Executable>
>> [2]
>> <http://cygwin.iphoneapps4free.org/release/apache2/apache2-2.2.6-1.tar.bz2>
>> [3] <http://en.wikipedia.org/wiki/Magic_number_(programming)>
>> [4] <http://www.delorie.com/djgpp/doc/coff/filhdr.html>
>>
>
>

For the archives:

>
> Hmmm, I think you should use i686-pc-cygwin. *That'll help make libtool
> and the like do what they do best on Cygwin.

Yes, that is absolutely necessary else bootstrappig doesn't work for
some packages.

>
> bootstrap-prefix.sh doesn't have support for this target (yet), since

Right.

> there is no profile for it either. *Though, I think it's going to harm
> you more to set a wrong CHOST than hacking your way through it.
>

You have to do.

Alternatively skip the tmp/ binaries use cygwins and fix the portage
settings in /etc. Use "portage/profiles/prefix" as default profile and
put below settings to /etc/make.conf. However I didn't test the
alternative way. It's theory.

> You'll have to create a profile for cygwin yourself (very limited, just
> containing the CHOST should do, keyword = x86-interix or something is an
> acceptable hack for now)

ARCH="x86"
CHOST="i686-pc-cygwin"
ACCEPT_KEYWORDS="-x86 x86-linux"

Al