[gentoo-dev] NonsolvableDepsInMisc

[Lucas Ramage]

Hello,

I submitted a pull request which adds new packages as dependencies for the
package I am attempting to maintain and I am receiving the following errors,

https://qa-reports.gentoo.org/output/gentoo-ci/aca3f2128/output.html#dev-python/paho-mqtt

NonsolvableDepsInDev

NonsolvableDepsInStable

​I have built the package in a container using FEATURE="test" and also ran
`repoman -dx full` as instructed in the Gentoo Github Documentation and I
had no problems.

-- 

[image: Visit online journal] 

*Lucas Ramage* / Software Engineer
ramage.lu...@openmailbox.org / (941) 404-6794

*PGP Fingerprint* / Learn More 
EAE7 45DF 818D 4948 DDA7 0F44 F52A 5A96 7B9B 6FB7


*Visit online journal*
http://lramage94.github.io 

[image: Github]  [image: Linkedin]

[gentoo-dev] Packages up for grabs: media-sound/lyvi

[Jonas Stein]

Dear all,

The following packages are up for grabs:

media-sound/lyvi

after retirement of the proxied maintainer.
https://packages.gentoo.org/packages/media-sound/lyvi

The package is already in EAPI=5 and in a good state without open bugs.

Lyvi is a lyrics, artist info and guitar tabs
viewer. On supported terminals, Lyvi can also show artist photos and
cover images.

Todo: Patch release; Lyvi is on the latest released bump, but there are
newer commits upstream.

-- 
Best,
Jonas







































signature.asc
Description: OpenPGP digital signature

[gentoo-dev] Getting glibc-2.26 stable

[Andreas K. Huettel]

Hey all, 

now that upstream has already released glibc-2.27, we should think about 
getting glibc-2.26 stable soon. That, however, means work for all of us.

Typical problems:

* Build failures because xlocale.h is gone
  https://bugs.gentoo.org/638010
  Fix: include locale.h instead

* Build failures, undefined reference to 'minor' / 'major'
  https://bugs.gentoo.org/575232
  Fix: additionally include sys/sysmacros.h
  No, we can't delay this further; the inclusion is (finally) 
  also upstream gone in 2.28. Everything we fix now will for
  sure save us trouble later.

* Build failures because of missing rpc support in glibc
  https://bugs.gentoo.org/381391
  See the following wiki page for information:
  https://wiki.gentoo.org/wiki/Project:Toolchain/RPC_implementation

* Build failures because of missing NIS support
  After fixing rpc support, add a dependency on net-libs/libnsl
  (unconditionally).

* Some internal type definitions have changed, which may lead
  to more obscure build problems (rare).

Please test and fix bugs!!! 

If you want to update your otherwise stable system, please
keyword the following (otherwise you'll get blockers):

~sys-libs/glibc-2.26
=net-libs/libnsl-1*
=net-libs/rpcsvc-proto-1*

This should be safe for amd64 and x86; for other arches it doesn't hurt to ask 
in #gentoo-toolchain or test in a chroot first. DON'T UPGRADE WITH MIPS OR 
SPARC YET.

Feel free to try emerging sys-libs/glibc with FEATURES=test 
(that takes a while). On amd64/x86, nearly all of the ~4000 tests succeed; I 
know of the following problems:
* elf/tst-prelink-cmp fails, still needs analysis
* only with gcc-7.3.0, six tests math/test-?float128* fail (with my CFLAGS)
  This is a gcc regression (miscompilation of truncf128), already submitted
  upstream.
In any case, I think this still looks nice:
Summary of test results:
  7 FAIL
   4109 PASS
 11 UNSUPPORTED
 29 XFAIL
  2 XPASS
Other arches may show more test failures; please file separate bugs per arch;
see https://bugs.gentoo.org/634988#c1 for what is needed for a useful report.

Stabilization tracker (fresh and largely yet empty): 
https://bugs.gentoo.org/show_bug.cgi?id=glibc-2.26-stable

General issue tracker: 
https://bugs.gentoo.org/show_bug.cgi?id=glibc-2.26

Cheers, 
Andreas

-- 
Andreas K. Hüttel
dilfri...@gentoo.org
Gentoo Linux developer
(council, toolchain, perl, libreoffice, comrel)

signature.asc
Description: This is a digitally signed message part.

Re: [gentoo-dev] RFC v2: eutils.eclass: More reliable return status for e*_clean.

[Michał Górny]

W dniu pią, 16.02.2018 o godzinie 19∶12 +0100, użytkownik Ulrich Mueller
napisał:
> > > > > > On Fri, 16 Feb 2018, Michał Górny wrote:
> > Could you please use quotes instead of backslash escapes all the
> > way? They're generally less confusing since escapes trigger
> > different behavior in different contexts in bash.
> 
> Done for the parentheses, and removed escaping of the plus sign which
> is no special character in bash.

Thanks. Looks good to me now.


-- 
Best regards,
Michał Górny

[gentoo-dev] RFC v2: eutils.eclass: More reliable return status for e*_clean.

[Ulrich Mueller]

> On Fri, 16 Feb 2018, Michał Górny wrote:

> Could you please use quotes instead of backslash escapes all the
> way? They're generally less confusing since escapes trigger
> different behavior in different contexts in bash.

Done for the parentheses, and removed escaping of the plus sign which
is no special character in bash.

> On Fri, 16 Feb 2018, Michael Orlitzky wrote:

> To make maintenance easier, I would go one step further and say that
> unless two functions need the same variables or call one another,
> they belong in separate eclasses. Since ecvs_clean, esvn_clean, and
> egit_clean are completely independent of one another, they could go
> in separate eclasses -- it's not like you'll need more than one of
> them in your ebuild. Then in the future if we need to change
> egit_clean, we will know precisely which ebuilds are affected.

If these were 200 line functions then I would agree. However, they
consist of one line each, are all about the same theme, and too much
fragmentation of eclasses should be avoided. In fact, I am already
reluctant about splitting off these tiny functions from eutils.

Anyway, updated patches are attached. First patch changes the code,
second patch splits the functions off into their own eclass (with no
code changes).

Ulrich

From 5cd1763df2dc385524da0163bdcadecfaf93675f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrich=20M=C3=BCller?= 
Date: Fri, 16 Feb 2018 07:47:46 +0100
Subject: [PATCH 1/2] eutils.eclass: More reliable return status for e*_clean
 functions.

In ecvs_clean, combine the two find commands into one, so that the
exit status of the first one won't be ignored.

Also use find -exec rather then find | xargs, so we don't have to
check the exit status of all commands in the pipeline.
---
 eclass/eutils.eclass | 8 
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/eclass/eutils.eclass b/eclass/eutils.eclass
index 8bbd561015ad..0a5bf3853582 100644
--- a/eclass/eutils.eclass
+++ b/eclass/eutils.eclass
@@ -44,8 +44,8 @@ fi
 # internal CVS directories.  Defaults to $PWD.
 ecvs_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name 'CVS' -prune -print0 | xargs -0 rm -rf
-   find "$@" -type f -name '.cvs*' -print0 | xargs -0 rm -rf
+   find "$@" '(' -type d -name 'CVS' -prune -o -type f -name '.cvs*' ')' \
+   -exec rm -rf '{}' +
 }
 
 # @FUNCTION: esvn_clean
@@ -55,7 +55,7 @@ ecvs_clean() {
 # internal Subversion directories.  Defaults to $PWD.
 esvn_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name '.svn' -prune -print0 | xargs -0 rm -rf
+   find "$@" -type d -name '.svn' -prune -exec rm -rf '{}' +
 }
 
 # @FUNCTION: egit_clean
@@ -65,7 +65,7 @@ esvn_clean() {
 # contains internal Git directories.  Defaults to $PWD.
 egit_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name '.git*' -prune -print0 | xargs -0 rm -rf
+   find "$@" -type d -name '.git*' -prune -exec rm -rf '{}' +
 }
 
 # @FUNCTION: emktemp
-- 
2.16.1

From 5056a7a455bac8a5b5bdadc357cc67cf0f6865fd Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrich=20M=C3=BCller?= 
Date: Fri, 16 Feb 2018 18:48:13 +0100
Subject: [PATCH 2/2] vcs-clean.eclass: Split off clean helpers from
 eutils.eclass.

Split off functions ecvs_clean, esvn_clean, and egit_clean into
a dedicated vcs-clean.eclass. No code changes.

For backwards compatibility, eutils inherits the new eclass in
existing EAPIs.
---
 eclass/eutils.eclass| 34 ++
 eclass/vcs-clean.eclass | 40 
 2 files changed, 42 insertions(+), 32 deletions(-)
 create mode 100644 eclass/vcs-clean.eclass

diff --git a/eclass/eutils.eclass b/eclass/eutils.eclass
index 0a5bf3853582..7840afbb77b9 100644
--- a/eclass/eutils.eclass
+++ b/eclass/eutils.eclass
@@ -20,7 +20,8 @@ _EUTILS_ECLASS=1
 # implicitly inherited (now split) eclasses
 case ${EAPI:-0} in
 0|1|2|3|4|5|6)
-   inherit desktop epatch estack ltprune multilib preserve-libs 
toolchain-funcs
+   inherit desktop epatch estack ltprune multilib preserve-libs \
+   toolchain-funcs vcs-clean
;;
 esac
 
@@ -37,37 +38,6 @@ if ! declare -F eqawarn >/dev/null ; then
}
 fi
 
-# @FUNCTION: ecvs_clean
-# @USAGE: [list of dirs]
-# @DESCRIPTION:
-# Remove CVS directories recursiveley.  Useful when a source tarball contains
-# internal CVS directories.  Defaults to $PWD.
-ecvs_clean() {
-   [[ $# -eq 0 ]] && set -- .
-   find "$@" '(' -type d -name 'CVS' -prune -o -type f -name '.cvs*' ')' \
-   -exec rm -rf '{}' +
-}
-
-# @FUNCTION: esvn_clean
-# @USAGE: [list of dirs]
-# @DESCRIPTION:
-# Remove .svn directories recursiveley.  Useful when a source tarball contains
-# internal Subversion directories.  Defaults to $PWD.
-esvn_clean() {
-   [[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name '.svn' -prune -exec rm -rf '{}' +
-}
-
-# @FUNCTION: egit_clean
-# @USAGE: [list of d

Re: [gentoo-dev] RFC: eutils.eclass: More reliable return status for e*_clean.

[Michał Górny]

W dniu pią, 16.02.2018 o godzinie 08∶06 -0500, użytkownik Michael
Orlitzky napisał:
> On 02/16/2018 03:46 AM, Ulrich Mueller wrote:
> > 
> > Should we take this as an opportunity to split off these three
> > functions into their own eclass, e.g. vcs-clean.eclass?
> 
> I think this is a good direction to go in. Changing a popular eclass is
> always scary, and the more unrelated stuff it contains, the harder it
> gets. It's not easy to tell which ebuilds use the part of the eclass
> that you're touching, so you wind up testing (or at least worrying
> about) them all. There's the metadata regen, too.
> 
> To make maintenance easier, I would go one step further and say that
> unless two functions need the same variables or call one another, they
> belong in separate eclasses. Since ecvs_clean, esvn_clean, and
> egit_clean are completely independent of one another, they could go in
> separate eclasses -- it's not like you'll need more than one of them in
> your ebuild. Then in the future if we need to change egit_clean, we will
> know precisely which ebuilds are affected.
> 

When you reach the point of having one eclass per one-command function,
you should have already figured out it's easier to inline the command
in ebuilds.

-- 
Best regards,
Michał Górny

Re: [gentoo-dev] RFC: Moving gnome2_icon_cache_update functionality to xdg-utils.eclass

[Mart Raudsepp]

On Fri, 2018-02-16 at 06:38 +0100, Michał Górny wrote:
> W dniu pią, 16.02.2018 o godzinie 02∶17 +0200, użytkownik Mart
> Raudsepp
> napisał:
> > Hello,
> > 
> > Before writing the obvious patches, would anyone terribly object to
> > moving gnome2_icon_cache_update into xdg-utils.eclass instead and
> > hooking it up via savelist guard in xdg.eclass (only running in
> > postinst/postrm when really needed)?
> > 
> > The gtk icon cache format isn't used by gtk+ only - it's used by Qt
> > as
> > well these days; and either way if anything installs things per the
> > icon specs, the caches should be updated for gtk/qt to see it (even
> > if
> > the program itself is written in motif or something).
> 
> Maybe it'd be reasonable to rename it while at it, to gtk_icon_cache*
> or xdg_gtk_icon_cache*. It would really suck to have xdg* provide
> a function named gnome2*, especially when it's not specific to GNOME.

Yes, of course. That's an obvious thing to do, didn't even think to
mention it explicitly. Afterall, we'll need the old name still in
gnome2-utils for compatibility for a while too.
I'm unsure about the name though, I like gtk_update_icon_cache, but not
having it prefixed with xdg or xdg-utils sounds a bit wrong; though I
think it makes a good argument to make an exception here. The tool that
provides the helper is split into a gtk-update-icon-cache named packge
as well.
Do you have any opinions on my dilemmas?

> > 
> > Open dilemmas:
> > 
> > * To DEPEND on gtk-update-icon-cache, or to rely on the toolkits to
> > pull it in and only run the updates if the helper is present. We
> > have
> > it split out from gtk+ codebase, so it only has glib/gdk-pixbuf
> > deps
> > right now.
> > 
> > * If it's worth trying to avoid double calls in postint and postrm
> > for
> > ebuilds that inherit xdg and call gnome2_icon_cache_update manually
> > right now on top (due to xdg.eclass not doing it for them).
> > 
> > 
> > If these changes are agreeable, then they could also be batched
> > into a
> > single push with the documentation update patches pending from
> > mgorny
> > for a single metadata cache invalidation of gnome2-utils.eclass
> > using
> > ebuilds.
> > 
> > If this looks good, I can write patches soon, though there's always
> > other things to do too (like work on integrating a gnome-
> > meson.eclass
> > or meson-desktop.eclass on top of this), so help with patches
> > welcome
> > as well.
> > 
> > 
> > Mart
> > 
> 
> 

Re: [gentoo-dev] RFC: eutils.eclass: More reliable return status for e*_clean.

[Michael Orlitzky]

On 02/16/2018 03:46 AM, Ulrich Mueller wrote:
>
> Should we take this as an opportunity to split off these three
> functions into their own eclass, e.g. vcs-clean.eclass?

I think this is a good direction to go in. Changing a popular eclass is
always scary, and the more unrelated stuff it contains, the harder it
gets. It's not easy to tell which ebuilds use the part of the eclass
that you're touching, so you wind up testing (or at least worrying
about) them all. There's the metadata regen, too.

To make maintenance easier, I would go one step further and say that
unless two functions need the same variables or call one another, they
belong in separate eclasses. Since ecvs_clean, esvn_clean, and
egit_clean are completely independent of one another, they could go in
separate eclasses -- it's not like you'll need more than one of them in
your ebuild. Then in the future if we need to change egit_clean, we will
know precisely which ebuilds are affected.

Re: [gentoo-dev] RFC: eutils.eclass: More reliable return status for e*_clean.

[Michał Górny]

Dnia 16 lutego 2018 09:46:33 CET, Ulrich Mueller  napisał(a):
>Currently the return status of e{cvs,svn,git}_clean is not at all
>reliable, because only the last command in the pipeline is tested.
>In addition, there are two pipelines in ecvs_clean, and the exit
>status of the first one is completely ignored.
>
>Another issue is that xargs -0 is an extension not specified by POSIX.
>See patch included below, fixing both problems. (Note that this is
>close to the implementation originally proposed in bug 210708 [1].)
>
>Should we take this as an opportunity to split off these three
>functions into their own eclass, e.g. vcs-clean.eclass?
>
>Ulrich
>
>[1] https://bugs.gentoo.org/210708
>
>
>From 83c0afbfb4ef0c501c65dcd8e0c9c8bd30cdd70a Mon Sep 17 00:00:00 2001
>From: =?UTF-8?q?Ulrich=20M=C3=BCller?= 
>Date: Fri, 16 Feb 2018 07:47:46 +0100
>Subject: [PATCH] eutils.eclass: More reliable return status for
>e*_clean.
>
>In ecvs_clean, combine the two find commands into one, so that the
>exit status of the first one won't be ignored.
>
>Also use find -exec rather then find | xargs, so we don't have to
>check the exit status of all commands in the pipeline.
>---
> eclass/eutils.eclass | 8 
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
>diff --git a/eclass/eutils.eclass b/eclass/eutils.eclass
>index 8bbd561015ad..bce8cf1c2610 100644
>--- a/eclass/eutils.eclass
>+++ b/eclass/eutils.eclass
>@@ -44,8 +44,8 @@ fi
> # internal CVS directories.  Defaults to $PWD.
> ecvs_clean() {
>   [[ $# -eq 0 ]] && set -- .
>-  find "$@" -type d -name 'CVS' -prune -print0 | xargs -0 rm -rf
>-  find "$@" -type f -name '.cvs*' -print0 | xargs -0 rm -rf
>+  find "$@" \( -type d -name 'CVS' -prune -o -type f -name '.cvs*' \) \
>+  -exec rm -rf '{}' \+

Could you please use quotes instead of backslash escapes all the way? They're 
generally less confusing since escapes trigger different behavior in different 
contexts in bash.

> }
> 
> # @FUNCTION: esvn_clean
>@@ -55,7 +55,7 @@ ecvs_clean() {
> # internal Subversion directories.  Defaults to $PWD.
> esvn_clean() {
>   [[ $# -eq 0 ]] && set -- .
>-  find "$@" -type d -name '.svn' -prune -print0 | xargs -0 rm -rf
>+  find "$@" -type d -name '.svn' -prune -exec rm -rf '{}' \+
> }
> 
> # @FUNCTION: egit_clean
>@@ -65,7 +65,7 @@ esvn_clean() {
> # contains internal Git directories.  Defaults to $PWD.
> egit_clean() {
>   [[ $# -eq 0 ]] && set -- .
>-  find "$@" -type d -name '.git*' -prune -print0 | xargs -0 rm -rf
>+  find "$@" -type d -name '.git*' -prune -exec rm -rf '{}' \+
> }
> 
> # @FUNCTION: emktemp


-- 
Best regards,
Michał Górny (by phone)

[gentoo-dev] RFC: eutils.eclass: More reliable return status for e*_clean.

[Ulrich Mueller]

Currently the return status of e{cvs,svn,git}_clean is not at all
reliable, because only the last command in the pipeline is tested.
In addition, there are two pipelines in ecvs_clean, and the exit
status of the first one is completely ignored.

Another issue is that xargs -0 is an extension not specified by POSIX.
See patch included below, fixing both problems. (Note that this is
close to the implementation originally proposed in bug 210708 [1].)

Should we take this as an opportunity to split off these three
functions into their own eclass, e.g. vcs-clean.eclass?

Ulrich

[1] https://bugs.gentoo.org/210708


From 83c0afbfb4ef0c501c65dcd8e0c9c8bd30cdd70a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrich=20M=C3=BCller?= 
Date: Fri, 16 Feb 2018 07:47:46 +0100
Subject: [PATCH] eutils.eclass: More reliable return status for e*_clean.

In ecvs_clean, combine the two find commands into one, so that the
exit status of the first one won't be ignored.

Also use find -exec rather then find | xargs, so we don't have to
check the exit status of all commands in the pipeline.
---
 eclass/eutils.eclass | 8 
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/eclass/eutils.eclass b/eclass/eutils.eclass
index 8bbd561015ad..bce8cf1c2610 100644
--- a/eclass/eutils.eclass
+++ b/eclass/eutils.eclass
@@ -44,8 +44,8 @@ fi
 # internal CVS directories.  Defaults to $PWD.
 ecvs_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name 'CVS' -prune -print0 | xargs -0 rm -rf
-   find "$@" -type f -name '.cvs*' -print0 | xargs -0 rm -rf
+   find "$@" \( -type d -name 'CVS' -prune -o -type f -name '.cvs*' \) \
+   -exec rm -rf '{}' \+
 }
 
 # @FUNCTION: esvn_clean
@@ -55,7 +55,7 @@ ecvs_clean() {
 # internal Subversion directories.  Defaults to $PWD.
 esvn_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name '.svn' -prune -print0 | xargs -0 rm -rf
+   find "$@" -type d -name '.svn' -prune -exec rm -rf '{}' \+
 }
 
 # @FUNCTION: egit_clean
@@ -65,7 +65,7 @@ esvn_clean() {
 # contains internal Git directories.  Defaults to $PWD.
 egit_clean() {
[[ $# -eq 0 ]] && set -- .
-   find "$@" -type d -name '.git*' -prune -print0 | xargs -0 rm -rf
+   find "$@" -type d -name '.git*' -prune -exec rm -rf '{}' \+
 }
 
 # @FUNCTION: emktemp
-- 
2.16.1


pgpHJCo2nLSnB.pgp
Description: PGP signature