tk-8.4.15-r1 on x86-macos

11-28-2007, 03:17 PM

Two issues with the tk-8.4.16-r1 ebuild on x86-macos:

(1) There is what looks like typo ("EPEFIX" instead of "EPREFIX").

(2) The adjustment of the CC_ and LD_SEARCH_FLAGS is unconditional,
whereas in the tcl ebuild it's conditional for darwin. I find I need
to have the SEARCH_FLAGS sed conditional in tk, too.

With attached diff, tk can be merged on x86-macos. Could you please
keyword it accordingly?

// Cheers; Johan

11-28-2007, 07:33 PM

On Wed, Nov 28, 2007 at 05:17:44PM +0100, Johan Hattne wrote:
> Two issues with the tk-8.4.16-r1 ebuild on x86-macos:
>
> (1) There is what looks like typo ("EPEFIX" instead of "EPREFIX").
>
> (2) The adjustment of the CC_ and LD_SEARCH_FLAGS is unconditional, whereas
> in the tcl ebuild it's conditional for darwin. I find I need to have the
> SEARCH_FLAGS sed conditional in tk, too.
>
> With attached diff, tk can be merged on x86-macos. Could you please
> keyword it accordingly?

Committed[1], thanks

-- Elias

[1] http://overlays.gentoo.org/proj/alt/changeset/13378

11-29-2007, 12:31 AM

On Nov 28, 2007, at 21:33 , Elias Pipping wrote:

On Wed, Nov 28, 2007 at 05:17:44PM +0100, Johan Hattne wrote:

Two issues with the tk-8.4.16-r1 ebuild on x86-macos:

(1) There is what looks like typo ("EPEFIX" instead of "EPREFIX").

(2) The adjustment of the CC_ and LD_SEARCH_FLAGS is
unconditional, whereas
in the tcl ebuild it's conditional for darwin. I find I need to
have the

SEARCH_FLAGS sed conditional in tk, too.

With attached diff, tk can be merged on x86-macos. Could you please
keyword it accordingly?

Committed[1], thanks

Thanks! The sed got a bit funny, though:

[[ ${CHOST} != *-darwin* ]] && sed blablabla || die

should be

[[ ${CHOST} != *-darwin* ]] && ( sed blablabla || die )

shouldn't it? Or maybe the if ... else is clearer?

// Cheers; Johan

--
gentoo-alt@gentoo.org mailing list

11-29-2007, 06:20 AM

On Thu, Nov 29, 2007 at 02:31:07AM +0100, Johan Hattne wrote:
>
> On Nov 28, 2007, at 21:33 , Elias Pipping wrote:
>
>> On Wed, Nov 28, 2007 at 05:17:44PM +0100, Johan Hattne wrote:
>>> Two issues with the tk-8.4.16-r1 ebuild on x86-macos:
>>>
>>> (1) There is what looks like typo ("EPEFIX" instead of "EPREFIX").
>>>
>>> (2) The adjustment of the CC_ and LD_SEARCH_FLAGS is unconditional,
>>> whereas
>>> in the tcl ebuild it's conditional for darwin. I find I need to have the
>>> SEARCH_FLAGS sed conditional in tk, too.
>>>
>>> With attached diff, tk can be merged on x86-macos. Could you please
>>> keyword it accordingly?
>>
>> Committed[1], thanks
>
> Thanks! The sed got a bit funny, though:
>
> [[ ${CHOST} != *-darwin* ]] && sed blablabla || die
>
> should be
>
> [[ ${CHOST} != *-darwin* ]] && ( sed blablabla || die )
>
> shouldn't it? Or maybe the if ... else is clearer?

No, that's perfectly fine

The if ... else construct was avoided to keep the difference to the
non-prefix tree small.

Also, there's no need for a sub-shell here. iirc, dying in a subshell
wouldn't even be fatal.

-- Elias

11-29-2007, 08:38 AM

On Nov 29, 2007, at 08:20 , Elias Pipping wrote:

On Thu, Nov 29, 2007 at 02:31:07AM +0100, Johan Hattne wrote:

Thanks! The sed got a bit funny, though:

[[ ${CHOST} != *-darwin* ]] && sed blablabla || die

should be

[[ ${CHOST} != *-darwin* ]] && ( sed blablabla || die )

shouldn't it? Or maybe the if ... else is clearer?

No, that's perfectly fine

I'm not so sure. AND takes precedence over OR, so

a && b || c

is equivalent to

( a && b ) || c

But in this case we want

a && ( b || c )

because we want c only if a is true but b is false.

// Cheers; Johan

--
gentoo-alt@gentoo.org mailing list

11-29-2007, 10:04 AM

On Thu, 2007-11-29 at 10:38 +0100, Johan Hattne wrote:

> I'm not so sure. AND takes precedence over OR, so

Nope, 'man bash' says:

Of these list operators, && and || have equal precedence,
followed by ; and &, which have equal precedence.

>
> a && b || c
>
> is equivalent to
>
> ( a && b ) || c

Yes, but

a || b && c

also is equivalent to

( a || b ) && c

In zlib, I've just seen a similar broken construct:

[[ ${CHOST} == *-darwin* ]]
|| chmod 755 libz.so.*
&& chmod 755 libz.*.dylib

On AIX fex I've just seen this one:

chmod: cannot access `libz.*.dylib': A file or directory in the path name does not exist.

So 'chmod dylib' is done if either
'[[ darwin ]]' is true
*or*
'chmod so' is true.

Fixed in r13382.

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

--
gentoo-alt@gentoo.org mailing list

11-29-2007, 10:42 AM

On Nov 29, 2007, at 12:04 , Michael Haubenwallner wrote:

On Thu, 2007-11-29 at 10:38 +0100, Johan Hattne wrote:

I'm not so sure. AND takes precedence over OR, so

Nope, 'man bash' says:

Of these list operators, && and || have equal precedence,
followed by ; and &, which have equal precedence.

Sorry, I should have checked that! My mistake!

a && b || c

is equivalent to

( a && b ) || c

Yes, but

a || b && c

also is equivalent to

( a || b ) && c

But I still believe that what we have tk is broken

[[ ${CHOST} != *-darwin* ]] && sed -i
-e "s,^(TK_CC_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/$
{mylibdir}',"
-e "s,^(TK_LD_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/$
{mylibdir}',"

"${ED}"/usr/${mylibdir}/tkConfig.sh || die

because it'll die if CHOST isn't darwin OR if the sed fails. In my
case CHOST is darwin.

// Cheers; Johan

--
gentoo-alt@gentoo.org mailing list

11-30-2007, 04:58 AM

On Thu, Nov 29, 2007 at 12:42:29PM +0100, Johan Hattne wrote:
>
> On Nov 29, 2007, at 12:04 , Michael Haubenwallner wrote:
>
>> On Thu, 2007-11-29 at 10:38 +0100, Johan Hattne wrote:
>>
>>> I'm not so sure. AND takes precedence over OR, so
>>
>> Nope, 'man bash' says:
>>
>> Of these list operators, && and || have equal precedence,
>> followed by ; and &, which have equal precedence.
>
> Sorry, I should have checked that! My mistake!
>
>>>
>>> a && b || c
>>>
>>> is equivalent to
>>>
>>> ( a && b ) || c
>>
>> Yes, but
>>
>> a || b && c
>>
>> also is equivalent to
>>
>> ( a || b ) && c
>>
>
> But I still believe that what we have tk is broken
>
> [[ ${CHOST} != *-darwin* ]] && sed -i
> -e
> "s,^(TK_CC_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
> -e
> "s,^(TK_LD_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
> "${ED}"/usr/${mylibdir}/tkConfig.sh || die
>
> because it'll die if CHOST isn't darwin OR if the sed fails. In my case
> CHOST is darwin.

That is correct.

Fixed in r13393[1]-13394[2]

-- Elias

http://overlays.gentoo.org/proj/alt/changeset/13393
http://overlays.gentoo.org/proj/alt/changeset/13394

11-30-2007, 07:25 AM

On Nov 30, 2007, at 06:58 , Elias Pipping wrote:

On Thu, Nov 29, 2007 at 12:42:29PM +0100, Johan Hattne wrote:

But I still believe that what we have tk is broken

[[ ${CHOST} != *-darwin* ]] && sed -i
-e
"s,^(TK_CC_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
-e
"s,^(TK_LD_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
"${ED}"/usr/${mylibdir}/tkConfig.sh || die

because it'll die if CHOST isn't darwin OR if the sed fails. In
my case

CHOST is darwin.

That is correct.

Fixed in r13393[1]-13394[2]

I'm happy! Thanks a lot!

// Cheers; Johan

--
gentoo-alt@gentoo.org mailing list

11-30-2007, 09:38 AM

On 30-11-2007 09:25:01 +0100, Johan Hattne wrote:
>>> [[ ${CHOST} != *-darwin* ]] && sed -i
>>> -e
>>> "s,^(TK_CC_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
>>> -e
>>> "s,^(TK_LD_SEARCH_FLAGS='.*)',1:${EPREFIX}/usr/${mylibdir}',"
>>> "${ED}"/usr/${mylibdir}/tkConfig.sh || die
>>>
>>> because it'll die if CHOST isn't darwin OR if the sed fails. In my case
>>> CHOST is darwin.
>>
>> That is correct.
>>
>> Fixed in r13393[1]-13394[2]
>
> I'm happy! Thanks a lot!

Blame it on me, by the way. Sorry. And thanks both for reporting and
fixing.

--
Fabian Groffen
Gentoo on a different level
--
gentoo-alt@gentoo.org mailing list