re: Right way to cope with user error in make test?

[Dan Kegel]

jwhite wrote:
> The attached patch is one approach - skip the tests in this case.
> Does this seem like a reasonable approach to folks?

I like it.

I would also like to see us get a lot tougher about
letting people build wine without a full complement
of development headers.  Maybe our configure
should have a sanity check at the end, and abort
if something optional but very common is missing;
we could add a --no-sanity-check option to enable
configuring in that case if it's really intended.
- Dan

Right way to cope with user error in make test?

[Austin English]

On Fri, May 16, 2008 at 2:05 PM, Dan Kegel  wrote:
> jwhite wrote:
> > The attached patch is one approach - skip the tests in this case.
> > Does this seem like a reasonable approach to folks?
>
> I like it.
>
> I would also like to see us get a lot tougher about
> letting people build wine without a full complement
> of development headers.  Maybe our configure
> should have a sanity check at the end, and abort
> if something optional but very common is missing;
> we could add a --no-sanity-check option to enable
> configuring in that case if it's really intended.
> - Dan

We could have dotests check the config.log to see if there were any
dependencies missing, and warn/fail as appropriate.

Right way to cope with user error in make test?

[Dan Kegel]

On Fri, May 16, 2008 at 12:54 PM, Austin English
 wrote:
> > I would also like to see us get a lot tougher about
> > letting people build wine without a full complement
> > of development headers.  Maybe our configure
> > should have a sanity check at the end, and abort
> > if something optional but very common is missing;
> > we could add a --no-sanity-check option to enable
> > configuring in that case if it's really intended.
>
> We could have dotests check the config.log to see if there were any
> dependencies missing, and warn/fail as appropriate.

That only helps dotests.  I'm thinking we should increase
build quality of wine for other situations, too.

Right way to cope with user error in make test?

[Austin English]

On Fri, May 16, 2008 at 2:57 PM, Dan Kegel  wrote:
> > We could have dotests check the config.log to see if there were any
> > dependencies missing, and warn/fail as appropriate.
>
> That only helps dotests.  I'm thinking we should increase
> build quality of wine for other situations, too.

True, but with Alexandre on vacation, you won't be able to fix
./configure now. Fixing it there is ideal, but if Alexandre won't
accept, at least we can get dotests better results.

Right way to cope with user error in make test?

[Vitaliy Margolen]

Dan Kegel wrote:
> jwhite wrote:
> > The attached patch is one approach - skip the tests in this case.
> > Does this seem like a reasonable approach to folks?
>
> I like it.
>
> I would also like to see us get a lot tougher about
> letting people build wine without a full complement
> of development headers.  Maybe our configure
> should have a sanity check at the end, and abort
> if something optional but very common is missing;
> we could add a --no-sanity-check option to enable
> configuring in that case if it's really intended.
> - Dan

IMHO configure should abort if any of the required libraries are missing and 
'--without-<bar>' flag was not given. Pretty much anything that prints out 
warning if it was not found.

Hopefully this will prevent broken builds and invalid bug/problem reports.

Vitaliy.