On 13/12/2018 12:07, Ian Abbott wrote:
If a local package "foo-mod" for a Linux kernel module is selected by a
local tristate configuration option PTXCONF_FOO_MOD, then based on the
template for Linux drivers, it would include the following in its
foo-mod.make file:

ifdef PTXCONF_FOO_MOD
$(STATEDIR)/kernel.targetinstall.post: $(STATEDIR)/foo-mod.targetinstall
endif

However, this rule will be invoked even if the PTXCONF_FOO_MOD option
has been configured with the tristate option 'm' and has not been
changed to 'y' by a selected_collectionconfig.  This causes the Linux
kernel module to be built and installed unintentionally.  To prevent
that happening, the above lines in the foo-mod.make file can be changed
as follows (using a conditional test suggested by Michael Olbrich):

ifneq ($(filter foo-mod,$(PTX_PACKAGES_SELECTED)),)
$(STATEDIR)/kernel.targetinstall.post: $(STATEDIR)/foo-mod.targetinstall
endif

This patch changes the template for Linux driver packages in accordance
with the above change.

Signed-off-by: Ian Abbott <[email protected]>
Cc: Michael Olbrich <[email protected]>
---
v2: Changed the conditional test to that suggested by Michael Olbrich in
Message-ID <[email protected]> to avoid
breaking the build if the package is configured using the tristate 'm'
option and no collection is selected.

  rules/templates/template-src-linux-driver-make | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rules/templates/template-src-linux-driver-make 
b/rules/templates/template-src-linux-driver-make
index e06a22497..c13857924 100644
--- a/rules/templates/template-src-linux-driver-make
+++ b/rules/templates/template-src-linux-driver-make
@@ -22,7 +22,7 @@ PACKAGES-$(PTXCONF_@PACKAGE@) += @package@
  @PACKAGE@_DIR         := $(BUILDDIR)/$(@PACKAGE@)
  @PACKAGE@_LICENSE     := unknown
-ifdef PTXCONF_@PACKAGE@
+ifneq ($(filter @package@,$(PTX_PACKAGES_SELECTED)),)
  $(STATEDIR)/kernel.targetinstall.post: $(STATEDIR)/@[email protected]
  endif

I know this is in 2019.01.0 now, but I don't think the new conditional test for adding this kernel.targetinstall.post dependency is actually working properly. I've had a few `ptxdist go` commands running with the -j N option, and quite often the kernel.targetinstall.post completes before the @[email protected], meaning that the module does not get installed to the ${PTXDIST_PLATFORMDIR}/root/lib/modules/$UNAME_RELEASE/extra as it should (where $UNAME_RELEASE is the kernel `uname -r` value).

The conditional test from v1 of the patch:

ifdef PTXCONF_@PACKAGE@)

still works for me, but will break the build for other people.

I'll try and work out why the new conditional test isn't working for me reliably (especially when building with parallel jobs), but I welcome any suggestions to try and fix it properly.

--
-=( Ian Abbott <[email protected]> || Web: www.mev.co.uk )=-
-=( MEV Ltd. is a company registered in England & Wales. )=-
-=( Registered number: 02862268.  Registered address:    )=-
-=( 15 West Park Road, Bramhall, STOCKPORT, SK7 3JZ, UK. )=-

_______________________________________________
ptxdist mailing list
[email protected]

Reply via email to