[gentoo-dev] Fixing eclass code relying on ${IUSE} greps?

[gentoo-dev] Fixing eclass code relying on ${IUSE} greps?

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 627 bytes --]

Hello,

As the 'has_iuse' thread worked out, right now PMS doesn't allow us to
grep IUSE for random values during runtime. Thus, all eclasses using
that to enable features per ebuild-defined IUSE are broken.

What way are we suggesting for fixing this? Unless we're going to
retroactively rewrite respective parts of the spec, we can't fix this
without changing API.

Honestly, I'm ready to remove all relevant code in
autotools-utils.eclass and fix all ebuilds in gx86 & Sunrise using it.
But if that's not the path we're going to take, I'd like to avoid
changing API then.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

[gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 568 bytes --]

Il giorno mer, 14/09/2011 alle 12.38 +0200, Michał Górny ha scritto:
> 
> Honestly, I'm ready to remove all relevant code in
> autotools-utils.eclass and fix all ebuilds in gx86 & Sunrise using it.
> But if that's not the path we're going to take, I'd like to avoid
> changing API then. 

If that means that you won't just add --enable-debug/--disable-debug on
ebuilds just because they declare IUSE=debug then please do so.

It's one of the worst ideas I have ever seen implemented :|

-- 
Diego Elio Pettenò — Flameeyes
http://blog.flameeyes.eu/

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Samuli Suominen]

On 09/14/2011 02:06 PM, Diego Elio Pettenò wrote:
> Il giorno mer, 14/09/2011 alle 12.38 +0200, Michał Górny ha scritto:
>>
>> Honestly, I'm ready to remove all relevant code in
>> autotools-utils.eclass and fix all ebuilds in gx86 & Sunrise using it.
>> But if that's not the path we're going to take, I'd like to avoid
>> changing API then. 
> 
> If that means that you won't just add --enable-debug/--disable-debug on
> ebuilds just because they declare IUSE=debug then please do so.
> 
> It's one of the worst ideas I have ever seen implemented :|
> 

I second that.  I've been "yelling" about it for years...

Same for the stupid assumption gnome2.eclass does with IUSE="doc" for
gtk-doc

[gentoo-dev] [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Michał Górny]

---
 eclass/autotools-utils.eclass |   13 +++++++++----
 1 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/eclass/autotools-utils.eclass b/eclass/autotools-utils.eclass
index 495244b..8731d6b 100644
--- a/eclass/autotools-utils.eclass
+++ b/eclass/autotools-utils.eclass
@@ -88,7 +88,7 @@ case ${EAPI:-0} in
 	*) die "EAPI=${EAPI} is not supported" ;;
 esac
 
-inherit autotools base
+inherit autotools base eutils
 
 EXPORT_FUNCTIONS src_prepare src_configure src_compile src_install src_test
 
@@ -220,8 +220,6 @@ autotools-utils_src_prepare() {
 # in myeconfargs are passed here to econf. Additionally following USE
 # flags are known:
 #
-# IUSE="debug" passes --disable-debug/--enable-debug to econf respectively.
-#
 # IUSE="static-libs" passes --enable-shared and either --disable-static/--enable-static
 # to econf respectively.
 autotools-utils_src_configure() {
@@ -232,7 +230,14 @@ autotools-utils_src_configure() {
 
 	# Handle debug found in IUSE
 	if has debug ${IUSE//+}; then
-		econfargs+=($(use_enable debug))
+		local debugarg=$(use_enable debug)
+		if ! has "${debugarg}" "${myeconfargs[@]}"; then
+			eqawarn 'Implicit $(use_enable debug) for IUSE="debug" is deprecated.'
+			eqawarn 'Please add the necessary arg to myeconfargs if requested.'
+			eqawarn 'The autotools-utils eclass will stop appending it on Oct 15th.'
+
+			econfargs+=("${debugarg}")
+		fi
 	fi
 
 	# Handle static-libs found in IUSE, disable them by default
-- 
1.7.6.1

[gentoo-dev] Re: [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 675 bytes --]

Il giorno mer, 14/09/2011 alle 17.23 +0200, Michał Górny ha scritto:
> 
> +                       eqawarn 'Implicit $(use_enable debug) for
> IUSE="debug" is deprecated.'
> +                       eqawarn 'Please add the necessary arg to
> myeconfargs if requested.'
> +                       eqawarn 'The autotools-utils eclass will stop
> appending it on Oct 15th.' 

Ehm, no. Just change all of it together and be done with it. If it
wasn't misused this would at worse cause some overlay to not build debug
properly; if it was misused then it would solve more issues than it
would cause.

-- 
Diego Elio Pettenò — Flameeyes
http://blog.flameeyes.eu/

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Re: [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 839 bytes --]

On Wed, 14 Sep 2011 17:56:05 +0200
Diego Elio Pettenò <flameeyes@gentoo.org> wrote:

> Il giorno mer, 14/09/2011 alle 17.23 +0200, Michał Górny ha scritto:
> > 
> > +                       eqawarn 'Implicit $(use_enable debug) for
> > IUSE="debug" is deprecated.'
> > +                       eqawarn 'Please add the necessary arg to
> > myeconfargs if requested.'
> > +                       eqawarn 'The autotools-utils eclass will
> > stop appending it on Oct 15th.' 
> 
> Ehm, no. Just change all of it together and be done with it. If it
> wasn't misused this would at worse cause some overlay to not build
> debug properly; if it was misused then it would solve more issues
> than it would cause.

Er; but then overlay authors won't even get a warning that the API
changed.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Re: [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 693 bytes --]

Il giorno mer, 14/09/2011 alle 18.16 +0200, Michał Górny ha scritto:
> 
> Er; but then overlay authors won't even get a warning that the API
> changed. 

Sorry I guess I misquoted.

Feel free to warn (for now) but don't append it anyway, otherwise the
eclass is simply not usable in some contexts. Unfortunately I've seen
too many packages out there where --disable-debug still enables it
(broken AC_ARG_ENABLE), or where --disable-debug stops _every_ symbol
emission (which is almost never what you want), or where --enable-debug
stops the whole optimisation (which again is almost never what you
want)....

-- 
Diego Elio Pettenò — Flameeyes
http://blog.flameeyes.eu/

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

[gentoo-dev] [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Michał Górny]

---
 eclass/autotools-utils.eclass |   11 +++++++----
 1 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/eclass/autotools-utils.eclass b/eclass/autotools-utils.eclass
index 495244b..9217d00 100644
--- a/eclass/autotools-utils.eclass
+++ b/eclass/autotools-utils.eclass
@@ -88,7 +88,7 @@ case ${EAPI:-0} in
 	*) die "EAPI=${EAPI} is not supported" ;;
 esac
 
-inherit autotools base
+inherit autotools base eutils
 
 EXPORT_FUNCTIONS src_prepare src_configure src_compile src_install src_test
 
@@ -220,8 +220,6 @@ autotools-utils_src_prepare() {
 # in myeconfargs are passed here to econf. Additionally following USE
 # flags are known:
 #
-# IUSE="debug" passes --disable-debug/--enable-debug to econf respectively.
-#
 # IUSE="static-libs" passes --enable-shared and either --disable-static/--enable-static
 # to econf respectively.
 autotools-utils_src_configure() {
@@ -232,7 +230,12 @@ autotools-utils_src_configure() {
 
 	# Handle debug found in IUSE
 	if has debug ${IUSE//+}; then
-		econfargs+=($(use_enable debug))
+		local debugarg=$(use_enable debug)
+		if ! has "${debugarg}" "${myeconfargs[@]}"; then
+			eqawarn 'Implicit $(use_enable debug) for IUSE="debug" is no longer supported.'
+			eqawarn 'Please add the necessary arg to myeconfargs if requested.'
+			eqawarn 'The autotools-utils eclass will stop warning about it on Oct 15th.'
+		fi
 	fi
 
 	# Handle static-libs found in IUSE, disable them by default
-- 
1.7.6.1

Re: [gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Nirbheek Chauhan]

On Wed, Sep 14, 2011 at 4:50 PM, Samuli Suominen <ssuominen@gentoo.org> wrote:
> I second that.  I've been "yelling" about it for years...
>
> Same for the stupid assumption gnome2.eclass does with IUSE="doc" for
> gtk-doc
>

For reference, ye olde bug: https://bugs.gentoo.org/show_bug.cgi?id=262491

-- 
~Nirbheek Chauhan

Gentoo GNOME+Mozilla Team

[gentoo-dev] Re: [PATCH autotools-utils] Deprecate implicit IUSE=debug.

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 208 bytes --]

Il giorno mer, 14/09/2011 alle 18.44 +0200, Michał Górny ha scritto:
> 
>  eclass/autotools-utils.eclass |   11 +++++++---- 

LGTM.

-- 
Diego Elio Pettenò — Flameeyes
http://blog.flameeyes.eu/

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Fixing eclass code relying on ${IUSE} greps?

[Mike Frysinger]

On Wed, Sep 14, 2011 at 06:38, Michał Górny wrote:
> As the 'has_iuse' thread worked out, right now PMS doesn't allow us to
> grep IUSE for random values during runtime. Thus, all eclasses using
> that to enable features per ebuild-defined IUSE are broken.

this statement isnt exactly clear.  no, eclasses cannot leverage IUSE
from ebuilds (and vice versa).  but eclasses/ebuilds can leverage
their own IUSE value.
-mike

Re: [gentoo-dev] Fixing eclass code relying on ${IUSE} greps?

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 897 bytes --]

On Wed, 14 Sep 2011 15:10:23 -0400
Mike Frysinger <vapier@gentoo.org> wrote:

> On Wed, Sep 14, 2011 at 06:38, Michał Górny wrote:
> > As the 'has_iuse' thread worked out, right now PMS doesn't allow us
> > to grep IUSE for random values during runtime. Thus, all eclasses
> > using that to enable features per ebuild-defined IUSE are broken.
> 
> this statement isnt exactly clear.  no, eclasses cannot leverage IUSE
> from ebuilds (and vice versa).  but eclasses/ebuilds can leverage
> their own IUSE value.

Erm, that's not really true. Consider that usually eclass functions are
called from ebuilds. At the time of calling, IUSE can basically contain
anything; if PM doesn't care much, that'd be probably last values set
-- ebuild values.

Remember than calling a sourced function does not re-evaluate the
eclass it was sourced from.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Fixing eclass code relying on ${IUSE} greps?

[Mike Frysinger]

On Wed, Sep 14, 2011 at 15:16, Michał Górny wrote:
> On Wed, 14 Sep 2011 15:10:23 -0400 Mike Frysinger wrote:
>> On Wed, Sep 14, 2011 at 06:38, Michał Górny wrote:
>> > As the 'has_iuse' thread worked out, right now PMS doesn't allow us
>> > to grep IUSE for random values during runtime. Thus, all eclasses
>> > using that to enable features per ebuild-defined IUSE are broken.
>>
>> this statement isnt exactly clear.  no, eclasses cannot leverage IUSE
>> from ebuilds (and vice versa).  but eclasses/ebuilds can leverage
>> their own IUSE value.
>
> Erm, that's not really true. Consider that usually eclass functions are
> called from ebuilds. At the time of calling, IUSE can basically contain
> anything; if PM doesn't care much, that'd be probably last values set
> -- ebuild values.

i was referring to global scope.  in src_* pkg_* funcs, the code
should see the merged value.  or maybe i havent read the spec close
enough.
-mike

Re: [gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 930 bytes --]

On Wed, 14 Sep 2011 13:06:23 +0200
Diego Elio Pettenò <flameeyes@gentoo.org> wrote:

> Il giorno mer, 14/09/2011 alle 12.38 +0200, Michał Górny ha scritto:
> > 
> > Honestly, I'm ready to remove all relevant code in
> > autotools-utils.eclass and fix all ebuilds in gx86 & Sunrise using
> > it. But if that's not the path we're going to take, I'd like to
> > avoid changing API then. 
> 
> If that means that you won't just add --enable-debug/--disable-debug
> on ebuilds just because they declare IUSE=debug then please do so.
> 
> It's one of the worst ideas I have ever seen implemented :|

Well, the other thing is IUSE=static-libs. I don't like it either but
this is probably a bigger case than the other.

The main resolution as I see it, is to simply drop IUSE=static-libs
from a lot of ebuilds where static libs aren't actually required by
anything or anyone.

-- 
Best regards,
Michał Górny

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 316 bytes --]

Re: [gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Diego Elio Pettenò]

[-- Attachment #1: Type: text/plain, Size: 950 bytes --]

Il giorno mer, 14/09/2011 alle 22.03 +0200, Michał Górny ha scritto:
> 
> The main resolution as I see it, is to simply drop IUSE=static-libs
> from a lot of ebuilds where static libs aren't actually required by
> anything or anyone. 

I would point at an old post of mine:

http://blog.flameeyes.eu/2009/09/28/removing-la-files-for-dum-w-uncertain-people

just to give an idea of where you can safely drop static-libs
altogether. Or to the most recent one on the topic:

http://blog.flameeyes.eu/2011/08/29/useless-flag-static-libs

that ignores one obvious case (the one where you have NO headers in
either the package or linked packages — the latter is the case for most
X11 libs nowadays).

And make sure you check against the version of  the package you're
adding static-libs to, before saying that it does install headers (or
the other way around).

-- 
Diego Elio Pettenò — Flameeyes
http://blog.flameeyes.eu/

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 490 bytes --]

Re: [gentoo-dev] Re: Fixing eclass code relying on ${IUSE} greps?

[Mike Frysinger]

On Wed, Sep 14, 2011 at 16:03, Michał Górny wrote:
> Well, the other thing is IUSE=static-libs. I don't like it either but
> this is probably a bigger case than the other.
>
> The main resolution as I see it, is to simply drop IUSE=static-libs
> from a lot of ebuilds where static libs aren't actually required by
> anything or anyone.

no ... the current policy (and i see no reason to change it) is:
 - always build shared and static libs
 - always build shared libs and control static libs via USE=static-libs

so if the library can be statically linked against, the ebuild needs
to support it.  a valid test is not "does any other package in the
tree want to link against this", nor is "i dont want static libs, and
i cant find anyone who does".  most people dont want to do static
linking most of the time, but that number is not 0.
-mike