On Wed, Feb 19, 2014 at 09:17:48AM -0500, Maupin, Chase wrote:
> >-----Original Message-----
> >From: Aravind Batni
> >Sent: Tuesday, February 18, 2014 7:41 PM
> >To: Dmytriyenko, Denys; Maupin, Chase
> >Cc: [email protected]
> >Subject: RE: [meta-arago] [PATCH] ti-rm: provides ti resouce
> >manager recipe for KeyStone devices
> >
> >Denys/Chase,
> >Thanks for the inputs - Please find the response in-lined.
> >
> >-Best Regards,
> >Aravind
> >
> >
> >> -----Original Message-----
> >> From: Dmytriyenko, Denys
> >> Sent: Friday, February 14, 2014 4:12 PM
> >> To: Maupin, Chase
> >> Cc: Aravind Batni; [email protected]
> >> Subject: Re: [meta-arago] [PATCH] ti-rm: provides ti resouce
> >manager recipe
> >> for KeyStone devices
> >>
> >> On Thu, Feb 13, 2014 at 02:14:56PM +0000, Maupin, Chase wrote:
> >> > >-----Original Message-----
> >> > >From: [email protected] [mailto:meta-
> >arago-
> >> > >[email protected]] On Behalf Of Aravind Batni
> >> > >Sent: Wednesday, February 12, 2014 4:15 PM
> >> > >To: [email protected]
> >> > >Cc: Aravind Batni
> >> > >Subject: [meta-arago] [PATCH] ti-rm: provides ti resouce
> >manager
> >> > >recipe for KeyStone devices
> >> > >
> >> > >- TI Resource Manager Low Level Driver
> >> > >
> >> > >Signed-off-by: Aravind Batni <[email protected]>
> >> > >---
> >> > > meta-arago-extras/recipes-bsp/ti-rm/ti-rm.inc | 7 ++++
> >> > > meta-arago-extras/recipes-bsp/ti-rm/ti-rm_git.bb | 45
> >> > >++++++++++++++++++++++
> >> > > 2 files changed, 52 insertions(+)
> >> > > create mode 100644 meta-arago-extras/recipes-bsp/ti-rm/ti-
> >rm.inc
> >> > > create mode 100644 meta-arago-extras/recipes-bsp/ti-rm/ti-
> >> > >rm_git.bb
> >> > >
> >> > >diff --git a/meta-arago-extras/recipes-bsp/ti-rm/ti-rm.inc
> >b/meta-
> >> > >arago-extras/recipes-bsp/ti-rm/ti-rm.inc
> >> > >new file mode 100644
> >> > >index 0000000..96da467
> >> > >--- /dev/null
> >> > >+++ b/meta-arago-extras/recipes-bsp/ti-rm/ti-rm.inc
> >> > >@@ -0,0 +1,7 @@
> >> > >+LICENSE = "TI BSD"
> >> > >+LIC_FILES_CHKSUM =
> >> > >"file://COPYING.txt;md5=dc61631b65360e6beb73b6c337800afc"
> >> > >+
> >> > >+BRANCH="master"
> >> > >+SRC_URI = "git://git.ti.com/keystone-rtos/rm-
> >> > >lld.git;protocol=git;branch=${BRANCH}"
> >> > >+# Below commit ID corresponds to DEV.RM_LLD.02.00.00.08
> >SRCREV =
> >> > >+"3a73cfe015214ff0401639f85fa5e52ea610e59d"
> >> >
> >> > If you aren't going to have multiple versions of a recipe the
> >.inc is
> >> > not required. If you do plan for multiple versions then the
> >SRCREV at
> >> > least is probably best left in the version specific recipe and
> >not in the .inc.
> >> > Looking at this I would think it likely that you could/should
> >just
> >> > roll this .inc into the regular .bb recipe.
> >> >
> >
> > [Aravind Batni] Yes, we can roll this to the regular .bb recipe
> >
> >> > >diff --git a/meta-arago-extras/recipes-bsp/ti-rm/ti-rm_git.bb
> >> >
> >> > Looking at the SRCREV below do you want to call this
> >02.00.00.08
> >> > version of the recipe instead of just _git?
> >> >
> >[Aravind Batni] We would like to call this as _git recipe since
> >we don't plan to provide multiple recipes per RM release.
> >
> >> > >b/meta-arago-extras/recipes-bsp/ti-rm/ti-rm_git.bb
> >> > >new file mode 100644
> >> > >index 0000000..7c8dad0
> >> > >--- /dev/null
> >> > >+++ b/meta-arago-extras/recipes-bsp/ti-rm/ti-rm_git.bb
> >> > >@@ -0,0 +1,45 @@
> >> > >+DESCRIPTION = "TI Resource Manager Low Level Driver"
> >> > >+
> >> > >+COMPATIBLE_MACHINE = "keystone"
> >> > >+
> >> > >+PR = "r0"
> >> >
> >> > You should probably set PV here if you are not going to change
> >this
> >> > recipe to a specific version.
> >
> >[Aravind Batni] Yes, I would add PV variable in the recipe
> >> >
> >> > >+DEPENDS="ti-ipc"
> >> > >+LLD-NAME="rm"
> >> > >+
> >> > >+include ti-rm.inc
> >> > >+
> >> > >+S = "${WORKDIR}/git"
> >> > >+LLD-BLD-DIR="${S}/ti/drv"
> >> > >+
> >> > >+PACKAGES =+ "${PN}-test"
> >> > >+
> >> > >+FILES_${PN}-test = "${bindir}/rmDspClientTest_*.out \
> >> > >+ ${bindir}/rmLinuxClientTest_*.out \
> >> > >+
> >${bindir}/ti/drv/rm/test/dts_files/*.dtb"
> >> > >+
> >> > >+do_configure () {
> >> > >+# tweak the directory structure to LLD way
> >> > >+ cd ${S}
> >> > >+ mkdir -p ${LLD-BLD-DIR}
> >> > >+ cd ${LLD-BLD-DIR}
> >> > >+ ln -s ${S} ${LLD-NAME}
> >>
> >> Also, please don't use dashes in variable names! Underscores,
> >while not
> >> recommended, are still acceptable:
> >>
> >> LLDBLDDIR - best from Bitbake perspective, not very human-
> >readable
> >> LLD_BLD_DIR - not perfect from Bitbake perspective, but
> >parseable and
> >> readable LLD-BLD-DIR - may cause all kinds of issues
> >>
> > [Aravind Batni] Yes, I would correct the recipe not to use any
> >dashes
> >>
> >> > I'm not sure I understand what you are trying to do here. It
> >seems
> >> > like you want ${WORKDIR}/git/ti/drv/rm to be pointed to ${S}?
> >Looking
> >> > below it seems like you then want to pass ${S}/ti/drm/rm,
> >which points
> >> > to ${S} to the make commands. So can't you just point things
> >to ${S}?
> >> >
> >> > Are you trying to work around the Makefile maybe looking for
> >other
> >> > files in the ti/drv directory? Since you created that
> >directory that
> >> > doesn't seem likely though. This seems like an issue best
> >handled by
> >> > updating the Makefiles to allow you to set paths and have a
> >set of
> >> > defaults. i.e. PATHX ?= "default path". That way you can
> >update PATHX as
> >> a parameter you pass.
> >> > But this seems strange to make new direcory structures that
> >then link
> >> > back to the base directory you were already in.
> >[Aravind Batni] The RM source does not permit to build from ${S}
> >directly. If we had a way to create ${S} (git clone) under
> >ti/drv/rm, then we don’t need to create the symbolic links to
> >build RM lld. This is done originally to support DSP builds that
> >are delivered from PDK, which demand this depth in the directory.
>
> OK, help me understand here. You are cloning to ${WORKDIR}/git, but you
> need the clone to really be in ${WORKDIR}/git/ti/drv/rm directory because
> the makefiles/sources demand this directory structure?
>
> Denys, are you aware of any option to control where the clone checks out to?
Yes, adding ;destsuffix=ti/drv/rm to SRC_URI will check out into that
directory relative to ${WORKDIR}. If you want the default "git" part
preserved, you can do ;destsuffix=git/ti/drv/rm - either way is fine.
> If the sources were stored in the repo under ti/drm/rm would this work?
>
>
> >> >
> >> > >+}
> >> > >+
> >> > >+do_compile () {
> >> > >+# Now build the lld in the updated directory
> >> > >+ cd ${LLD-BLD-DIR}/${LLD-NAME}
> >> > >+ make -f makefile_armv7 clean
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2h
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME}
> >> > >+ make -f makefile_armv7 lib tests
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2h
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME}
> >> > >+ make -f makefile_armv7 lib tests
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2h
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME} USEDYNAMIC_LIB=yes
> >> > >+ make -f makefile_armv7 clean
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2k
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME}
> >> > >+ make -f makefile_armv7 lib tests
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2k
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME}
> >> > >+ make -f makefile_armv7 lib tests
> >> > >PDK_INSTALL_PATH=${STAGING_INCDIR} DEVICE=k2k
> >> RM_SRC_DIR=${LLD-
> >> > >BLD-DIR}/${LLD-NAME} USEDYNAMIC_LIB=yes
> >> >
> >> > Some thoughts:
> >> >
> >> > 1. Would this be better done as a for loop iterated of the
> >different
> >> > DEVICE settings?
> >[Aravind Batni] Yes, thanks for the inputs. I will modify the
> >recipe to have for loops .
> >> > 2. Since you don't seem to be breaking these libraries out per
> >DEVICE
> >> > and I think you are packaging both static and dynamic
> >libraries would an
> >> "all"
> >> > make target be better than calling each individually?
> >> > - I actually wonder if you would prefer to split dynamic vs.
> >static
> >> > libraries. Why are both packaged? Or maybe I don't
> >understand
> >> what
> >> > you are doing here?
> >> > 3. Should the libraries be packaged per DEVICE? The root of
> >this
> >> > question is whether this recipe should be machine specific and
> >you
> >> > build the package for k2k and k2h devices. It seems like you
> >are
> >> > making one package that has support for multiple devices.
> >
> >[Aravind Batni] There is a single library for both K2H and K2K
> >devices, the options are for test applications. I now modified the
> >recipe
> >to build the library only once for both K2H and K2K and build all
> >other variations on the test applications in the next updated
> >patch submit.
> >> >
> >> > >+}
> >> > >+
> >> > >+do_install () {
> >> > >+ install -d ${D}/${includedir}/ti/drv/${LLD-NAME}
> >> > >+ install -d ${D}/${libdir}
> >> > >+ install -d ${D}/${bindir}
> >> > >+ make -f makefile_armv7 install installbin
> >installbin_test
> >> > >INSTALL_INC_BASE_DIR=${D}/${includedir}
> >> > >INSTALL_LIB_BASE_DIR=${D}/${libdir}
> >> > >INSTALL_BIN_BASE_DIR=${D}/${bindir} DEVICE=k2h
> >> > >+ make -f makefile_armv7 install installbin
> >installbin_test
> >> > >INSTALL_INC_BASE_DIR=${D}/${includedir}
> >> > >INSTALL_LIB_BASE_DIR=${D}/${libdir}
> >> > >INSTALL_BIN_BASE_DIR=${D}/${bindir} DEVICE=k2k
> >> > >+}
> >> > >--
> >[Aravind Batni] I would have a for loop based installs for every
> >device.
> >> > >1.7.9.5
> >> > >
> >> > >_______________________________________________
> >> > >meta-arago mailing list
> >> > >[email protected]
> >> > >http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
> >> > _______________________________________________
> >> > meta-arago mailing list
> >> > [email protected]
> >> > http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
_______________________________________________
meta-arago mailing list
[email protected]
http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
