Bug 507086 – media-libs/libechonest unnecessary dependency on dev-qt/qttest:4

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Julian Ospald 2014-04-07 19:45:05 UTC]

It's a build system bug.

dev-qt/qttest is not needed if we do not build the tests

also, the ebuild should probably use the cmake option:
option(ECHONEST_BUILD_TESTS "Build all unit tests" ON)

in one way or another

[Comment 1 Julian Ospald 2014-04-07 19:47:30 UTC]

that means I tested the build with qttest uninstalled and
sed -i -e '/find_package/s/QtTest//' CMakeLists.txt || die

[Comment 2 Samuli Suominen (RETIRED) 2014-04-07 20:02:22 UTC]

Nikoli, OK to drop the dep and push the sed line in? I trust you test this ;)

[Comment 3 Nikoli 2014-04-07 20:05:22 UTC]

> also, the ebuild should probably use the cmake option:
option(ECHONEST_BUILD_TESTS "Build all unit tests" ON)

libechonest-2.2.0-r1.ebuild already has -DECHONEST_BUILD_TESTS=OFF, will check if qttest is required.

[Comment 4 Julian Ospald 2014-04-07 20:33:48 UTC]

(In reply to Nikoli from comment #3)
> > also, the ebuild should probably use the cmake option:
> option(ECHONEST_BUILD_TESTS "Build all unit tests" ON)
> 
> libechonest-2.2.0-r1.ebuild already has -DECHONEST_BUILD_TESTS=OFF, will
> check if qttest is required.

my tree must be out of date then

[Comment 5 Julian Ospald 2014-04-07 21:46:07 UTC]

btw. it's not clear to me why the tests are restricted

the tests of curl also need a network connection afais... I personally wouldn't make that a blocker

[Comment 6 Nikoli 2014-04-08 08:02:03 UTC]

> btw. it's not clear to me why the tests are restricted

The general rule is too disable tests which fail without connection to internet or with FEATURES=network-sandbox

> the tests of curl also need a network connection afais...

and curl ebuild has RESTRICT="test"

[Comment 7 Julian Ospald 2014-04-08 12:56:47 UTC]

(In reply to Nikoli from comment #6)
> and curl ebuild has RESTRICT="test"

Yes, but not because of that reason.

[Comment 8 Julian Ospald 2014-04-08 13:01:46 UTC]

see bug 285669 for an explanation

[Comment 9 Samuli Suominen (RETIRED) 2014-04-08 13:05:34 UTC]

Nikoli, OK to commit following?

--- libechonest-2.2.0-r1.ebuild
+++ libechonest-2.2.0-r1.ebuild
@@ -19,8 +19,7 @@
 RDEPEND=">=dev-libs/qjson-0.5
 	dev-qt/qtcore:4"
 DEPEND="${RDEPEND}
-	virtual/pkgconfig
-	dev-qt/qttest:4"
+	virtual/pkgconfig"
 
 DOCS=( AUTHORS README TODO )

[Comment 10 Julian Ospald 2014-04-08 13:10:08 UTC]

(In reply to Samuli Suominen from comment #9)
> Nikoli, OK to commit following?
> 
> --- libechonest-2.2.0-r1.ebuild
> +++ libechonest-2.2.0-r1.ebuild
> @@ -19,8 +19,7 @@
>  RDEPEND=">=dev-libs/qjson-0.5
>  	dev-qt/qtcore:4"
>  DEPEND="${RDEPEND}
> -	virtual/pkgconfig
> -	dev-qt/qttest:4"
> +	virtual/pkgconfig"
>  
>  DOCS=( AUTHORS README TODO )

That might fail without the sed (at least for 2.1.0).

[Comment 11 Nikoli 2014-04-08 13:30:33 UTC]

I prefer to wait until patch for this bug is accepted upstream.

[Comment 12 Julian Ospald 2014-04-08 14:14:49 UTC]

(In reply to Nikoli from comment #11)
> I prefer to wait until patch for this bug is accepted upstream.

fixing build systems shouldn't need upstream approval... it's trivial enough

[Comment 13 Nikoli 2014-04-08 14:21:47 UTC]

Even if it is trivial, it is not urgent - nothing is broken or insecure.

[Comment 14 Julian Ospald 2014-04-08 15:13:10 UTC]

(In reply to Nikoli from comment #13)
> Even if it is trivial, it is not urgent - nothing is broken or insecure.

I fail to see an argument.

[Comment 15 Julian Ospald 2014-04-08 15:21:15 UTC]

I would not get anything done in gentoo if I would follow that policy for trivial build system fixes...

[Comment 16 Samuli Suominen (RETIRED) 2014-04-08 16:20:07 UTC]

Long as it's reported upstream, the fix can be pushed to Portage, or especially then it can be pushed to Portage

So I've committed it to 2.2.0-r1