| Message ID | 1381463805-10521-1-git-send-email-mattst88@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Oct 10, 2013 at 08:56:45PM -0700, Matt Turner wrote: > We were seriously *requiring* libdrm_nouveau unless explicitly disabled? Acked-by: Ben Widawsky <ben@bwidawsk.net> > --- > configure.ac | 7 +++++-- > 1 file changed, 5 insertions(+), 2 deletions(-) > > diff --git a/configure.ac b/configure.ac > index f65942f..43740f9 100644 > --- a/configure.ac > +++ b/configure.ac > @@ -92,8 +92,11 @@ AM_CONDITIONAL(BUILD_ASSEMBLER, [test "x$enable_assembler" = xyes]) > # ----------------------------------------------------------------------------- > # for dma-buf tests > AC_ARG_ENABLE(nouveau, AS_HELP_STRING([--disable-nouveau], > - [Enable use of nouveau API for prime tests (default: enabled)]), > - [NOUVEAU=$enableval], [NOUVEAU=yes]) > + [Enable use of nouveau API for prime tests (default: auto)]), > + [NOUVEAU=$enableval], [NOUVEAU=auto]) > +if test "x$NOUVEAU" = xauto; then > + PKG_CHECK_EXISTS([libdrm_nouveau >= 2.4.33], [NOUVEAU=yes], [NOUVEAU=no]) > +fi > if test "x$NOUVEAU" = xyes; then > PKG_CHECK_MODULES(DRM_NOUVEAU, [libdrm_nouveau >= 2.4.33]) > AC_DEFINE(HAVE_NOUVEAU, 1, [Have nouveau support]) > -- > 1.8.3.2 >
On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote:
> We were seriously *requiring* libdrm_nouveau unless explicitly disabled?
I've had a bit of hilarious fail with optional testcases that
automatically get disabled when depencies aren't around. Hence why
they're all required by default with the optional switch to disable
them.
-Daniel
On Fri, Oct 11, 2013 at 12:40 AM, Daniel Vetter <daniel@ffwll.ch> wrote: > On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote: >> We were seriously *requiring* libdrm_nouveau unless explicitly disabled? > > I've had a bit of hilarious fail with optional testcases that > automatically get disabled when depencies aren't around. Hence why > they're all required by default with the optional switch to disable > them. You've had this problem in general, or with nouveau? I mean, this is kind of how autotools is supposed to work. If you explicitly want to ensure something is built you add --enable-thing.
On Fri, Oct 11, 2013 at 12:51:34PM -0700, Matt Turner wrote: > On Fri, Oct 11, 2013 at 12:40 AM, Daniel Vetter <daniel@ffwll.ch> wrote: > > On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote: > >> We were seriously *requiring* libdrm_nouveau unless explicitly disabled? > > > > I've had a bit of hilarious fail with optional testcases that > > automatically get disabled when depencies aren't around. Hence why > > they're all required by default with the optional switch to disable > > them. > > You've had this problem in general, or with nouveau? In general. > I mean, this is kind of how autotools is supposed to work. If you > explicitly want to ensure something is built you add --enable-thing. Yeah, but otoh I'm positively incompetent ;-) And since the main use case for i-g-t is to distribute the tests autoconf doesn't quite work like it usually does. -Daniel
On Fri, Oct 11, 2013 at 2:24 PM, Daniel Vetter <daniel@ffwll.ch> wrote: > On Fri, Oct 11, 2013 at 12:51:34PM -0700, Matt Turner wrote: >> On Fri, Oct 11, 2013 at 12:40 AM, Daniel Vetter <daniel@ffwll.ch> wrote: >> > On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote: >> >> We were seriously *requiring* libdrm_nouveau unless explicitly disabled? >> > >> > I've had a bit of hilarious fail with optional testcases that >> > automatically get disabled when depencies aren't around. Hence why >> > they're all required by default with the optional switch to disable >> > them. >> >> You've had this problem in general, or with nouveau? > > In general. So I think what I'm saying is that the nouveau support in intel-gpu-tools isn't the project's primary purpose, so we should allow it to build without nouveau support if you don't have nouveau installed. If you don't have nouveau installed, you're not making a mistake by not building intel-gpu-tools without support. And if you have nouveau installed, you still get nouveau support.
On 10/11/2013 02:29 PM, Matt Turner wrote: > On Fri, Oct 11, 2013 at 2:24 PM, Daniel Vetter <daniel@ffwll.ch> wrote: >> On Fri, Oct 11, 2013 at 12:51:34PM -0700, Matt Turner wrote: >>> On Fri, Oct 11, 2013 at 12:40 AM, Daniel Vetter <daniel@ffwll.ch> wrote: >>>> On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote: >>>>> We were seriously *requiring* libdrm_nouveau unless explicitly disabled? >>>> >>>> I've had a bit of hilarious fail with optional testcases that >>>> automatically get disabled when depencies aren't around. Hence why >>>> they're all required by default with the optional switch to disable >>>> them. >>> >>> You've had this problem in general, or with nouveau? >> >> In general. > > So I think what I'm saying is that the nouveau support in > intel-gpu-tools isn't the project's primary purpose, so we should > allow it to build without nouveau support if you don't have nouveau > installed. If you don't have nouveau installed, you're not making a > mistake by not building intel-gpu-tools without support. And if you > have nouveau installed, you still get nouveau support. I completely agree with Matt. If the system doesn't already have nouveau installed, then it there is very low likelihood that it makes any sense to install nouveau onto that system. It very likely does not have NVidia hardware at all. So why require it? Let's just autodetect it and behave sanely. Requiring libdrm_nouveau is a nuisance for developers working on platforms where libdrm doesn't get built with NVidia support.
On 10/11/2013 03:51 PM, Chad Versace wrote: > On 10/11/2013 02:29 PM, Matt Turner wrote: >> On Fri, Oct 11, 2013 at 2:24 PM, Daniel Vetter <daniel@ffwll.ch> wrote: >>> On Fri, Oct 11, 2013 at 12:51:34PM -0700, Matt Turner wrote: >>>> On Fri, Oct 11, 2013 at 12:40 AM, Daniel Vetter <daniel@ffwll.ch> wrote: >>>>> On Fri, Oct 11, 2013 at 5:56 AM, Matt Turner <mattst88@gmail.com> wrote: >>>>>> We were seriously *requiring* libdrm_nouveau unless explicitly disabled? >>>>> >>>>> I've had a bit of hilarious fail with optional testcases that >>>>> automatically get disabled when depencies aren't around. Hence why >>>>> they're all required by default with the optional switch to disable >>>>> them. >>>> >>>> You've had this problem in general, or with nouveau? >>> >>> In general. >> >> So I think what I'm saying is that the nouveau support in >> intel-gpu-tools isn't the project's primary purpose, so we should >> allow it to build without nouveau support if you don't have nouveau >> installed. If you don't have nouveau installed, you're not making a >> mistake by not building intel-gpu-tools without support. And if you >> have nouveau installed, you still get nouveau support. > > I completely agree with Matt. > > If the system doesn't already have nouveau installed, then it there is very > low likelihood that it makes any sense to install nouveau onto that system. > It very likely does not have NVidia hardware at all. So why require it? > Let's just autodetect it and behave sanely. > > Requiring libdrm_nouveau is a nuisance for developers working on platforms > where libdrm doesn't get built with NVidia support. And, this patch gets Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
diff --git a/configure.ac b/configure.ac index f65942f..43740f9 100644 --- a/configure.ac +++ b/configure.ac @@ -92,8 +92,11 @@ AM_CONDITIONAL(BUILD_ASSEMBLER, [test "x$enable_assembler" = xyes]) # ----------------------------------------------------------------------------- # for dma-buf tests AC_ARG_ENABLE(nouveau, AS_HELP_STRING([--disable-nouveau], - [Enable use of nouveau API for prime tests (default: enabled)]), - [NOUVEAU=$enableval], [NOUVEAU=yes]) + [Enable use of nouveau API for prime tests (default: auto)]), + [NOUVEAU=$enableval], [NOUVEAU=auto]) +if test "x$NOUVEAU" = xauto; then + PKG_CHECK_EXISTS([libdrm_nouveau >= 2.4.33], [NOUVEAU=yes], [NOUVEAU=no]) +fi if test "x$NOUVEAU" = xyes; then PKG_CHECK_MODULES(DRM_NOUVEAU, [libdrm_nouveau >= 2.4.33]) AC_DEFINE(HAVE_NOUVEAU, 1, [Have nouveau support])