[Buildroot] [PATCHv4] toolchain: fix installing gconv libs with multi-arch toolchain

[Buildroot] [PATCHv4] toolchain: fix installing gconv libs with multi-arch toolchain

[Yann E. MORIN]

For a multi-arch toolchain, gconv modules are in a sub-directory named
after the machine gcc targets. This is the case, for example, for the
Linaro ARM 2014.09 toolchain, which has the gconv modules in (relative
to the sysroot):
    /usr/lib/arm-linux-gnueabihf/gconv

while the Sourcery CodeBench ARM 2014.05 (non-multi-arch) has them in:
    /usr/lib/gconv

So, to catter for both cases, search both paths. We want to favour the
machine-specific gconv modules over potentially existing "generic" ones,
so we first search that (if it exists) and fallback to looking in the
generic location.

Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>

---
changes v3 -> v4:
  - nothing, just rebased on top of master

Changes v2 -> v3:
  - use TOOLCHAIN_EXTERNAL_PREFIX instead of computing it again

Changes v1 -> v2:
  - also handle the case where the gconv modules list is specified
---
 toolchain/toolchain.mk | 26 ++++++++++++++++++--------
 1 file changed, 18 insertions(+), 8 deletions(-)

diff --git a/toolchain/toolchain.mk b/toolchain/toolchain.mk
index 3f9900b..0a35909 100644
--- a/toolchain/toolchain.mk
+++ b/toolchain/toolchain.mk
@@ -17,28 +17,38 @@ endif
 ifeq ($(BR2_TOOLCHAIN_GLIBC_GCONV_LIBS_COPY),y)
 GCONV_LIBS = $(call qstrip,$(BR2_TOOLCHAIN_GLIBC_GCONV_LIBS_LIST))
 define COPY_GCONV_LIBS
-	$(Q)if [ -z "$(GCONV_LIBS)" ]; then \
-		$(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/gconv/gconv-modules \
-				      $(TARGET_DIR)/usr/lib/gconv/gconv-modules; \
-		$(INSTALL) -m 0644 $(STAGING_DIR)/usr/lib/gconv/*.so \
+	$(Q)found_gconv=no; \
+	for d in $(TOOLCHAIN_EXTERNAL_PREFIX) ''; do \
+		[ -d "$(STAGING_DIR)/usr/lib/$${d}/gconv" ] || continue; \
+		found_gconv=yes; \
+		break; \
+	done; \
+	if [ "$${found_gconv}" = "no" ]; then \
+		printf "Unable to find gconv modules\n" >&2; \
+		exit 1; \
+	fi; \
+	if [ -z "$(GCONV_LIBS)" ]; then \
+		$(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/$${d}/gconv/gconv-modules \
+				      $(TARGET_DIR)/usr/lib/gconv/gconv-modules && \
+		$(INSTALL) -m 0644 $(STAGING_DIR)/usr/lib/$${d}/gconv/*.so \
 				   $(TARGET_DIR)/usr/lib/gconv \
 		|| exit 1; \
 	else \
 		for l in $(GCONV_LIBS); do \
-			$(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/gconv/$${l}.so \
+			$(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/$${d}/gconv/$${l}.so \
 					      $(TARGET_DIR)/usr/lib/gconv/$${l}.so \
 			|| exit 1; \
-			$(TARGET_READELF) -d $(STAGING_DIR)/usr/lib/gconv/$${l}.so |\
+			$(TARGET_READELF) -d $(STAGING_DIR)/usr/lib/$${d}/gconv/$${l}.so |\
 			sort -u |\
 			sed -e '/.*(NEEDED).*\[\(.*\.so\)\]$$/!d; s//\1/;' |\
 			while read lib; do \
-				 $(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/gconv/$${lib} \
+				 $(INSTALL) -m 0644 -D $(STAGING_DIR)/usr/lib/$${d}/gconv/$${lib} \
 						       $(TARGET_DIR)/usr/lib/gconv/$${lib} \
 				 || exit 1; \
 			done; \
 		done; \
 		./support/scripts/expunge-gconv-modules "$(GCONV_LIBS)" \
-			<$(STAGING_DIR)/usr/lib/gconv/gconv-modules \
+			<$(STAGING_DIR)/usr/lib/$${d}/gconv/gconv-modules \
 			>$(TARGET_DIR)/usr/lib/gconv/gconv-modules; \
 	fi
 endef
-- 
1.9.1

[Buildroot] [PATCHv4] toolchain: fix installing gconv libs with multi-arch toolchain

[Thomas Petazzoni]

Dear Yann E. MORIN,

On Tue, 21 Apr 2015 18:55:52 +0200, Yann E. MORIN wrote:
> For a multi-arch toolchain, gconv modules are in a sub-directory named
> after the machine gcc targets. This is the case, for example, for the
> Linaro ARM 2014.09 toolchain, which has the gconv modules in (relative
> to the sysroot):
>     /usr/lib/arm-linux-gnueabihf/gconv
> 
> while the Sourcery CodeBench ARM 2014.05 (non-multi-arch) has them in:
>     /usr/lib/gconv
> 
> So, to catter for both cases, search both paths. We want to favour the
> machine-specific gconv modules over potentially existing "generic" ones,
> so we first search that (if it exists) and fallback to looking in the
> generic location.
> 
> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> 
> ---
> changes v3 -> v4:
>   - nothing, just rebased on top of master

I've applied, even though I'm not terribly happy with the huge amount
of complex shell code being added. This gconv stuff is already calling
a shell script to recreate the gconv-modules file. Maybe we should just
move all this logic to a separate shell script, in which we could use
functions and so on?

I've tried to think a bit about writing all of this in make, but it
isn't trivial at all.

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

[Buildroot] [PATCHv4] toolchain: fix installing gconv libs with multi-arch toolchain

[Yann E. MORIN]

Thomas, All,

On 2015-04-22 22:48 +0200, Thomas Petazzoni spake thusly:
> On Tue, 21 Apr 2015 18:55:52 +0200, Yann E. MORIN wrote:
> > For a multi-arch toolchain, gconv modules are in a sub-directory named
> > after the machine gcc targets. This is the case, for example, for the
> > Linaro ARM 2014.09 toolchain, which has the gconv modules in (relative
> > to the sysroot):
> >     /usr/lib/arm-linux-gnueabihf/gconv
> > 
> > while the Sourcery CodeBench ARM 2014.05 (non-multi-arch) has them in:
> >     /usr/lib/gconv
> > 
> > So, to catter for both cases, search both paths. We want to favour the
> > machine-specific gconv modules over potentially existing "generic" ones,
> > so we first search that (if it exists) and fallback to looking in the
> > generic location.
> > 
> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
> > 
> > ---
> > changes v3 -> v4:
> >   - nothing, just rebased on top of master
> 
> I've applied, even though I'm not terribly happy with the huge amount
> of complex shell code being added. This gconv stuff is already calling
> a shell script to recreate the gconv-modules file. Maybe we should just
> move all this logic to a separate shell script, in which we could use
> functions and so on?

Yes, that's a good idea. I've put it on my ever-growing TODO-list. ;-)

> I've tried to think a bit about writing all of this in make, but it
> isn't trivial at all.

Nah, it's not. I tried a bit too, but promptly aborted, to keep using
shell stuff.

Thanks! :-)

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'