Hi,

I'm currently working on packaging dtrace which doesn't (at least
currently, may not ever) support running tests as non-root, but
does support handling installing them for later manual use.

This raises a question: how should we control installing such tests? How
should the user request that?

USE=test isn't suitable because:
a) I generally expect it not to mutate the image;
b) ago, rightly, based on this has a tinderbox check which looks for added/removed files when tests are enabled.

I'd like to pick some name which is suitable for us to use elsewhere and
it's not really a package-specific issue. Thoughts?

Please assume the tests are large enough to justify conditional install
and have additional dependencies.

(This is arguably related to https://bugs.gentoo.org/867799 and so on
wrt dev-lang/python always installing tests.)

thanks,
sam

-----BEGIN PGP SIGNATURE-----

iOUEARYKAI0WIQQlpruI3Zt2TGtVQcJzhAn1IN+RkAUCZrY5JV8UgAAAAAAuAChp c3N1ZXItZnByQG5vdGF0aW9ucy5vcGVucGdwLmZpZnRoaG9yc2VtYW4ubmV0MjVB NkJCODhERDlCNzY0QzZCNTU0MUMyNzM4NDA5RjUyMERGOTE5MA8cc2FtQGdlbnRv by5vcmcACgkQc4QJ9SDfkZDczgD9Ekx1XYYPIjmPVv9fuNZzj7wgSaI53MAFXaVg I01GuG0A/j/SXwGGos75kT/Mw5w8mCQDX31H3y8lOs6ud0G6qlwL
=PR5W
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

To: gentoo-dev@lists.gentoo.org
To: sam@gentoo.org (Sam James)
Copy: arsen@gentoo.org (=?UTF-8?Q?Arsen_Arsenovi=C4=87?=)
Copy: python@gentoo.org

This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------eT8N82E4p62B9AoYCYA0nSyg
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On 8/9/24 11:59 AM, Maciej Barć wrote:

Hi Sam,

We have a "source" flag so maybe we could either use that or have "test-source" BUT we also have FEATURES="installsources" ...so (if I understand that feature correctly) we can just run test from /usr/src/debug/${CATEGORY}/${PF}

But IMO I'd rather clone the project and run test from the clone.

Of source the drawback of installsources & git cloning is that emerge
does not run any modifications on the tests nor applies patches to them
nor buildsystem/testrunner used.

installsources is not the same thing at all. That parses ELF files to
extract .c / .cpp etc source filenames from debuginfo to save them.

This thread is asking about the idea of compiling test programs and
installing the test executables so that you can run them later. You
absolutely cannot do that with a directory of .c files, for numerous
reasons including the fact that it only saves sources for files which
get installed anyway, but also because it does not install build system
files (meson.build, Makefile, CMakeLists.txt, etc).

P.S. See my gnome documentation link, it talks about git cloning too.

--
Eli Schwartz


--------------eT8N82E4p62B9AoYCYA0nSyg--

-----BEGIN PGP SIGNATURE-----

wnsEABYIACMWIQTnFNnmK0TPZHnXm3qEp9ErcA0vVwUCZrY+SAUDAAAAAAAKCRCEp9ErcA0vVx8b AP91nworiXzI5o2+2WjTg0S40Q+XAh1iq1+IAT3mUqhrggEArhpGTMtfX9tzMcioCApWGHgQiE1E ti/BRUF20RuXTw0=
=m/6G
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

To: sam@gentoo.org (Sam James)
Copy: arsen@gentoo.org (=?UTF-8?Q?Arsen_Arsenovi=C4=87?=)
Copy: python@gentoo.org

This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------6n7DU8JM00tIsIwDeXRrbmwQ
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

On 8/9/24 11:43 AM, Sam James wrote:

Hi,

I'm currently working on packaging dtrace which doesn't (at least
currently, may not ever) support running tests as non-root, but
does support handling installing them for later manual use.

This raises a question: how should we control installing such tests? How should the user request that?

USE=test isn't suitable because:
a) I generally expect it not to mutate the image;
b) ago, rightly, based on this has a tinderbox check which looks for added/removed files when tests are enabled.

I'd like to pick some name which is suitable for us to use elsewhere and
it's not really a package-specific issue. Thoughts?

Please assume the tests are large enough to justify conditional install
and have additional dependencies.

(This is arguably related to https://bugs.gentoo.org/867799 and so on
wrt dev-lang/python always installing tests.)

This reminds me of
https://wiki.gnome.org/Initiatives/GnomeGoals/InstalledTests so it's
definitely not a dtrace specific issue.


--
Eli Schwartz


--------------6n7DU8JM00tIsIwDeXRrbmwQ--

-----BEGIN PGP SIGNATURE-----

wnsEABYIACMWIQTnFNnmK0TPZHnXm3qEp9ErcA0vVwUCZrY77QUDAAAAAAAKCRCEp9ErcA0vVwqS AP9dubOuVsGfDxcefB5nkTqcM1XGiWbl1AhBdIhyF/+fTQD+LvtP9zcKr6oHAVUb0TNJHQSY26GB P7QJCF/Rlos1EA4=
=+flw
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Fri, 2024-08-09 at 16:43 +0100, Sam James wrote:

Hi,

I'm currently working on packaging dtrace which doesn't (at least
currently, may not ever) support running tests as non-root, but
does support handling installing them for later manual use.

This raises a question: how should we control installing such tests? How should the user request that?

USE=test isn't suitable because:
a) I generally expect it not to mutate the image;
b) ago, rightly, based on this has a tinderbox check which looks for added/removed files when tests are enabled.

I'd like to pick some name which is suitable for us to use elsewhere and
it's not really a package-specific issue. Thoughts?

Please assume the tests are large enough to justify conditional install
and have additional dependencies.

Sounds a bit related to USE=examples.

--
Best regards,
Michał Górny


-----BEGIN PGP SIGNATURE-----

iQFGBAABCgAwFiEEx2qEUJQJjSjMiybFY5ra4jKeJA4FAma2QAMSHG1nb3JueUBn ZW50b28ub3JnAAoJEGOa2uIyniQOLOwIAM1yohxVzpzjXDSvctR6+D1qguSAfFlJ 8JZgHWVVdfLQU8MiRhypDU8UAI5bOYJdsHa6DTWXLd1PF1edkrjXIoKe1QW6NpT1 CENZi1Z+KunWOkXFc0gha/CtK74Y4d1sxOwLGGIQ5TJ3jPTba7tzuLEpbdnpvQRQ ZXvjNckHibptVJd5j5aJZkwgBKca6YabUaxw6hB6iGNJl16P0raX6+q2QqiBXwlq vK2XIkiikFAsKjIaWc+Jmq5d697pZk6ciMPv9o/Cpce+OQ0ylgDdh8ULDXYI0teu ZykxklQ/NxMMACtkHdCAg6IP2i2+Ths4dZsMAwVDjDdHz2hn1tFlsKw=
=uayW
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------qV50G4FM3A1HohPsUzSas8vM
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Hey,

On 9.8.2024 18.43, Sam James wrote:

I'd like to pick some name which is suitable for us to use elsewhere and
it's not really a package-specific issue. Thoughts?

the first simple thought that came to my mind: "install-tests" or "install-test-files". Both are available and should be pretty
self-explanatory.

-- juippis

--------------qV50G4FM3A1HohPsUzSas8vM--

-----BEGIN PGP SIGNATURE-----

iQGTBAEBCgB9FiEEltRJ9L6XRmDQCngHc4OUK43AaWIFAma+3VNfFIAAAAAALgAo aXNzdWVyLWZwckBub3RhdGlvbnMub3BlbnBncC5maWZ0aGhvcnNlbWFuLm5ldDk2 RDQ0OUY0QkU5NzQ2NjBEMDBBNzgwNzczODM5NDJCOERDMDY5NjIACgkQc4OUK43A aWKVtgf+NJ9LvcBgp59t8MQ+gBcG9L9p0XmRoxZb9K2xsJkmvtsb8IgCpbowpBJa LVySmnG7s+2UjgOL3bXZcCyozPbtz5iek+aRabxFsyz56czY0mswmIfE2L0otPBv Vwaa4cP97qAdAR/P06TSGk90TemHs/13d/J/d5/U3JgFQA6P43RLAIgjfUHUfB2K exuu/Ok1fdnw5Jo2CBOgCoZ1QHjQEj16DaHR2sSP64tOgam3UfiyAWN4aK9j8bPR lDvtNGGlb4B7RK/yynqcJXN2s6v+NcgDjCfPA/4czhnjDI1olIZzy2ufmTMcsQO3 MOKJsZ9mIChXCNDCHf+gd0yqG+f5IA==
=DcGR
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Joonas Niilola <juippis@gentoo.org> writes:

Hey,

On 9.8.2024 18.43, Sam James wrote:

I'd like to pick some name which is suitable for us to use elsewhere and
it's not really a package-specific issue. Thoughts?

the first simple thought that came to my mind: "install-tests" or "install-test-files". Both are available and should be pretty self-explanatory.

Thoughts on prefix/suffix?

"test-install", "test-install-files" are more in-keeping with USE=test, USE=test-full, etc, but maybe that's *bad* because it's not related
at all to what the ebuild does, so we should do install-* like you've suggested?

It's been in the back of my head so wanted to bring it up before
committing to install-*. I think it's the better option indeed but want
to check.

-- juippis

thanks,
sam

--=-=-Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----

iOUEARYKAI0WIQQlpruI3Zt2TGtVQcJzhAn1IN+RkAUCZsJGR18UgAAAAAAuAChp c3N1ZXItZnByQG5vdGF0aW9ucy5vcGVucGdwLmZpZnRoaG9yc2VtYW4ubmV0MjVB NkJCODhERDlCNzY0QzZCNTU0MUMyNzM4NDA5RjUyMERGOTE5MA8cc2FtQGdlbnRv by5vcmcACgkQc4QJ9SDfkZB/UAD+KqsW0hEmFdWsRNfm3Z1Ca5ySaJIVXDuBdNVP nSp8kPUBAPoHUO3wYvn61vl7TkHJbXD4BK0tK6Z6LMb0aXa/a9UG
=/VLX
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Copy: gentoo-dev@lists.gentoo.org

This is an OpenPGP/MIME signed message (RFC 4880 and 3156) --------------0v1TAug0QfqFAl20Yy6f08a5
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Hey,

On 18.8.2024 22.06, Sam James wrote:

Thoughts on prefix/suffix?

"test-install", "test-install-files" are more in-keeping with USE=test, USE=test-full, etc, but maybe that's *bad* because it's not related
at all to what the ebuild does, so we should do install-* like you've suggested?

My instant thought was that "install*" prefix is more accurate and
easier to understand. But "test*" makes it more consistent with other test-related stuff. No hard opinion either way, I'd prefer "install*"
just because it's clearer to me.

-- juippis

--------------0v1TAug0QfqFAl20Yy6f08a5--

-----BEGIN PGP SIGNATURE-----

iQGTBAEBCgB9FiEEltRJ9L6XRmDQCngHc4OUK43AaWIFAmbC0TZfFIAAAAAALgAo aXNzdWVyLWZwckBub3RhdGlvbnMub3BlbnBncC5maWZ0aGhvcnNlbWFuLm5ldDk2 RDQ0OUY0QkU5NzQ2NjBEMDBBNzgwNzczODM5NDJCOERDMDY5NjIACgkQc4OUK43A aWJn2QgAsyP79O2jChSEUc2YuR7C1sMfg4rDJY0Of10UZGdo2+EullgBNRy4Hqj2 kJj6KbmKKS58OGqmq9kclez6X3EoaUgyz+H5yAnjPHn1nwvR4Lh66jyRYPaiRkod xBHCOhsD21xzAIOqj1RhAmnlZSsuK6LjFUuu8w9xchNzXyAucWfumi5y+XPLqRAp vjoNFU2s5X3wXeJM99tMmeSQaO9pwfm5mx8dUr5jqW3iU8Zj5QKgAc1vHDLbVSX+ CkorxR9O8QTnGCViE+UhdPJUgHPa9msUADT9BQr83vT1KeskavLkcYoh7Mm7L1EL RS4b/6d5e3TzFBBUJm4oplyG7h/C/w==
=QyER
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)