Skipping tests during install/build

Skipping tests during install/build

[Mike Gerwitz]

[-- Attachment #1: Type: text/plain, Size: 1318 bytes --]

Hello, everyone:

Is there a way (without screwing anything up) to skip tests during a
build?  I understand that this is generally unwise---I don't want to
debate those merits.

My immediate problem is that I'm on a dinky little ARM C201 Chromebook
and any sort of building is quite time-consuming, and often
prohibitively so: I can deal with the compilation times, but the tests
are simply too much; I don't have time to wait potentially hours for
software to build if they aren't available from hydra.  GnuTLS is one
particularly intense dependency test-wise, for example.  And then if a
test fails for whatever reason, I'm completely out of luck.  I'd rather
install and then run tests later at my leisure, accepting the risks.

But I don't know if any test output is taken into account in any Guix
hashes.

There are a few situations where I've had no choice but to fall back to
installing the respective Debian package(s).  But I've been very
impressed with how many ARM packages _are_ available from hydra---many
more than I had expected!

Anyway: any suggestions? :/

-- 
Mike Gerwitz
Free Software Hacker+Activist | GNU Maintainer & Volunteer
GPG: D6E9 B930 028A 6C38 F43B  2388 FEF6 3574 5E6F 6D05
Old: 2217 5B02 E626 BC98 D7C0  C2E5 F22B B815 8EE3 0EAB
https://mikegerwitz.com

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]

Re: Skipping tests during install/build

[Ludovic Courtès]

Hi Mike,

Mike Gerwitz <mtg@gnu.org> skribis:

> Is there a way (without screwing anything up) to skip tests during a
> build?  I understand that this is generally unwise---I don't want to
> debate those merits.

The short answer is “no”.  The Boolean that determines whether tests are
run is an “input” of the build process, and thus it contributes to that
/gnu/store hash.  Changing it leads to a different hash.

I think it’s a feature, though.  :-)

> My immediate problem is that I'm on a dinky little ARM C201 Chromebook
> and any sort of building is quite time-consuming, and often
> prohibitively so: I can deal with the compilation times, but the tests
> are simply too much; I don't have time to wait potentially hours for
> software to build if they aren't available from hydra.  GnuTLS is one
> particularly intense dependency test-wise, for example.  And then if a
> test fails for whatever reason, I'm completely out of luck.  I'd rather
> install and then run tests later at my leisure, accepting the risks.
>
> But I don't know if any test output is taken into account in any Guix
> hashes.
>
> There are a few situations where I've had no choice but to fall back to
> installing the respective Debian package(s).  But I've been very
> impressed with how many ARM packages _are_ available from hydra---many
> more than I had expected!

The intent is to have as much as possible available as substitutes.
However, while this works well for x86_64, the other platforms are not
in as good a state.

Part of it is due to the fact that they have fewer build machines¹ so
they tend to lag behind.  The second problem is that they have fewer
users; developers don’t always notice when something breaks there and
cannot fix issues easily if they don’t have access to the hardware.

So I think it’s a chicken-and-egg problem.  Reporting the problems that
you have on ARM (test suite failures, build failures, etc.) can help
raise awareness and get people to fix things more quickly.

Thanks,
Ludo’.

¹ See the list at <https://www.gnu.org/software/guix/donate/>.

Re: Skipping tests during install/build

[Mike Gerwitz]

[-- Attachment #1: Type: text/plain, Size: 1288 bytes --]

On Sun, Feb 12, 2017 at 14:30:01 +0100, Ludovic Courtès wrote:
> The short answer is “no”.  The Boolean that determines whether tests are
> run is an “input” of the build process, and thus it contributes to that
> /gnu/store hash.  Changing it leads to a different hash.

I was hoping for a `guix package -i --dont-do-this` kind of flag, not
changing the inputs.

> I think it’s a feature, though.  :-)

It is.  I'm just looking to selfishly subvert your good intentions. :)

> The intent is to have as much as possible available as substitutes.
> However, while this works well for x86_64, the other platforms are not
> in as good a state.

Oh, I want to be clear that I don't have a problem with the state of ARM
substitutes in itself; like I said, I'm surprised there are as many as
there are.

> So I think it’s a chicken-and-egg problem.  Reporting the problems that
> you have on ARM (test suite failures, build failures, etc.) can help
> raise awareness and get people to fix things more quickly.

Yep, will do.

Thanks.

-- 
Mike Gerwitz
Free Software Hacker+Activist | GNU Maintainer & Volunteer
GPG: D6E9 B930 028A 6C38 F43B  2388 FEF6 3574 5E6F 6D05
Old: 2217 5B02 E626 BC98 D7C0  C2E5 F22B B815 8EE3 0EAB
https://mikegerwitz.com

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 818 bytes --]