On Wed, Jun 07, 2017 at 05:16:18PM +0530, Shreyansh Jain wrote:
> On Wednesday 07 June 2017 03:58 PM, Hunt, David wrote:
> > Hi Shreyansh,
> >
> >
> > On 7/6/2017 10:36 AM, Shreyansh Jain wrote:
> > > Hello David,
> > >
> > > On Wednesday 07 June 2017 02:09 PM, Hunt, David wrote:
> > > > Shreyansh,
> > > >
> > > > I found an issue (or two) with this part of the patch, and
> > > > have a proposed solution.
> > > >
> > > > 1. RTE_TARGET originally had a different meaning. It was used
> > > > for making examples, specifying the target directory of where
> > > > the SDK was built. It's not good to re-purpose this for
> > > > something else, as I'm doing in this patch. (even though I'm not
> > > > sure that variable is suitably named in the first place, but
> > > > that's a different issue).
> > >
> > > Even I didn't realize this until you highlighted here.
> > >
> > > > 2. If we set RTE_TARGET on the environment, we will break the
> > > > 'make -C examples/<app>', unless we set RTE_TARGET to be
> > > > something else (i.e. 'make -C examples/<app> RTE_TARGET=build').
> > > > One value for making DPDK, and another for building examples.
> > > > It's confusing to the user.
> > >
> > > Agree about re-using RTE_TARGET is breaking existing assumption about
> > > its use.
> > >
> > > >
> > > > An alternative patch would be as follows:
> > > >
> > > > RTE_CONFIG_TEMPLATE :=
> > > > ifdef T
> > > > *-ifeq ("$(origin T)", "command line")*
> > > > RTE_CONFIG_TEMPLATE := $(RTE_SRCDIR)/config/defconfig_$(T)
> > > > *-endif**
> > > > *endif
> > > > export RTE_CONFIG_TEMPLATE
> > > So, that would mean, user would do either of the following:
> > >
> > > make T=<template> config
> > >
> > > or
> > >
> > > export T=<template>
> > > make config
> > >
> > > Is that correct? (I tried it and it seems to be working fine)
> > > First method is same as today. For the second, I am just skeptical
> > > whether we should use such a small identifier ("T") or we have a new
> > > RTE_TEMPLATE.
> > >
> > > Either way, I am OK. [export T=<template>] looks fine to me - in fact,
> > > on a second though, IMO, if T=<template> is provided as command
> > > line, it should also be acceptable as env variable.
> > >
> >
> > I did a quick poll here in the office and people feel that 'T' is too
> > short for an environment variable. RTE_TEMPLATE would be preferred, and
> > it's a sensible choice that does not conflict with RTE_TARGET.
> >
> > So if we use RTE_TEMPLATE, we'd also have to put in a couple of lines
> > for the "make install" case, but it's still a small enough patch:
> >
> > diff --git a/mk/rte.sdkinstall.mk b/mk/rte.sdkinstall.mk
> > index dbac2a2..a464b01 100644
> > --- a/mk/rte.sdkinstall.mk
> > +++ b/mk/rte.sdkinstall.mk
> > @@ -47,6 +47,10 @@ ifneq ($(MAKECMDGOALS),pre_install)
> > include $(RTE_SDK)/mk/rte.vars.mk
> > endif
> >
> > *+ifndef T**
> > **+T := $(RTE_TEMPLATE)**
> > **+endif**
> > * ifdef T # defaults with T= will install an almost flat staging tree
> > export prefix ?=
> > kerneldir ?= $(prefix)/kmod
> >
> >
> > diff --git a/mk/rte.sdkroot.mk b/mk/rte.sdkroot.mk
> > index 076a2d7..0b71a4e 100644
> > --- a/mk/rte.sdkroot.mk
> > +++ b/mk/rte.sdkroot.mk
> > @@ -63,6 +63,8 @@ ifdef T
> > ifeq ("$(origin T)", "command line")
> > RTE_CONFIG_TEMPLATE := $(RTE_SRCDIR)/config/defconfig_$(T)
> > endif
> > *+else**
> > **+RTE_CONFIG_TEMPLATE := $(RTE_SRCDIR)/config/defconfig_$(RTE_TEMPLATE)**
> > * endif
> > export RTE_CONFIG_TEMPLATE
> >
> > So if T is provided on the command line, it takes priority.
> > If that seems reasonable to you, I'll push up a v3. :)
>
> Sounds good to me.
> Feel free to add my signoff to v3.
>
> >
In another shameless plug, can you guys perhaps take a look at the RFC
I've just posted[1], as an option to move away from this whole area of
RTE_TARGET/RTE_SDK, and the complexity such as this that it brings. Let
me know any thoughts or comments you have.
Thanks,
/Bruce
[1] http://dpdk.org/ml/archives/dev/2017-June/067428.html
