base.bbclass: add make-native to default dependencies
RFC 1/1 base.bbclass: add make-native to default dependencies
In case the build system is running some ancient make, we can just as easily build our own make (3.82) to get predictable behaviour. Does something like this make sense?
classes/base.bbclass +1 -1
@@ -71,7 +71,7 @@ def base_deps(d): if not bb.data.getVar('INHIBIT_DEFAULT_DEPS', d): if (bb.data.getVar('HOST_SYS', d, 1) != bb.data.getVar('BUILD_SYS', d, 1)): - deps += " virtual/${TARGET_PREFIX}gcc virtual/libc " + deps += " make-native virtual/${TARGET_PREFIX}gcc virtual/libc " elif bb.data.inherits_class('native', d) and \ bb.data.getVar('PN', d, True) not in \ ("linux-libc-headers-native",
"quilt-native",
Quoted 3 lines · Michael Smith
In case the build system is running some ancient make, we can just as easily build our own make (3.82) to get predictable behaviour. Does something like this make sense?
although it does but make 3.82 is not usable out of box yet. so people might find surprises. 3.81 would be ok.
Quoted 21 lines
classes/base.bbclass +0
-0@@ -71,7 +71,7 @@ def base_deps(d):if not
bb.data.getVar('INHIBIT_DEFAULT_DEPS', d):
if (bb.data.getVar('HOST_SYS', d, 1)!=bb.data.getVar('BUILD_SYS', d, 1)):
- deps += " virtual/${TARGET_PREFIX}gcc virtual/libc "
+ deps += " make-native virtual/${TARGET_PREFIX}gcc virtual/libc "
elifbb.data.inherits_class('native', d)and \
bb.data.getVar('PN', d, True) not in \
("linux-libc-headers-native", "quilt-native",_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
Signature · 1 line
-Khem
Quoted 2 lines · Khem Raj
although it does but make 3.82 is not usable out of box yet. so people might find surprises. 3.81 would be ok.
I could add a DEFAULT_PREFERENCE = "-1" to the 3.82 recipe and mention the packages that don't build with it in a comment. Do you remember which packages fail to build with 3.82?
Quoted 14 lines
Quoted 14 lines
classes/base.bbclass +1
-1@@ -71,7 +71,7 @@ def base_deps(d): if not bb.data.getVar('INHIBIT_DEFAULT_DEPS', d): if (bb.data.getVar('HOST_SYS', d, 1) != bb.data.getVar('BUILD_SYS', d, 1)): - deps += " virtual/${TARGET_PREFIX}gcc virtual/libc " + deps += " make-native virtual/${TARGET_PREFIX}gccvirtual/libc "
elif bb.data.inherits_class('native', d) and \ bb.data.getVar('PN', d, True) not in \ ("linux-libc-headers-native", "quilt-native",
Quoted 5 lines · Michael Smith
Quoted 2 lines · Khem Raj
although it does but make 3.82 is not usable out of box yet. so people might find surprises. 3.81 would be ok.
I hit eglibc there may be more.
Quoted 27 lines
I could add a
DEFAULT_PREFERENCE= "-1" to the 3.82 recipe and mention the packages that don't build with it in a comment. Do you remember which packages fail to build with 3.82?Quoted 8 lines
Quoted 8 lines
classes/base.bbclass +0
-0@@ -71,7 +71,7 @@ def base_deps(d):if not
bb.data.getVar('INHIBIT_DEFAULT_DEPS', d):
if (bb.data.getVar('HOST_SYS', d, 1)!=bb.data.getVar('BUILD_SYS', d, 1)):Quoted 8 lines
Quoted 8 lines
- deps += " virtual/${TARGET_PREFIX}gcc virtual/libc "
+ deps += " make-native virtual/${TARGET_PREFIX}gcc virtual/libc "
elifbb.data.inherits_class('native', d)and \
bb.data.getVar('PN', d, True)not in \
("linux-libc-headers-native", "quilt-native",_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
Signature · 1 line
-Khem
Quoted 7 lines · Khem Raj
Quoted 4 lines · Michael Smith
Quoted 2 lines · Khem Raj
although it does but make 3.82 is not usable out of box yet. so people might find surprises. 3.81 would be ok.
I hit eglibc there may be more.
Hmm, OK. The patch below wouldn't affect eglibc's dependencies, so we'd still have to change eglibc.inc and eglibc-initial.inc to depend on make-native, or people with make 3.82 on their host system will be unable to build eglibc.
Or, apply a patch to every version of eglibc to fix the build with make 3.82.
Which do you prefer? I can only really test glibc 2.9, but adding make-native to the DEPENDS in glibc.inc/glibc-initial.inc should work about the same as doing it in eglibc.
Mike
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
classes/base.bbclass +0
-1@@ -71,7 +71,7 @@ def base_deps(d): if not bb.data.getVar('INHIBIT_DEFAULT_DEPS', d): if (bb.data.getVar('HOST_SYS', d, 1) != bb.data.getVar('BUILD_SYS', d, 1)): - deps += " virtual/${TARGET_PREFIX}gccvirtual/libc "
+ deps += " make-native virtual/${TARGET_PREFIX}gcc
virtual/libc "elif bb.data.inherits_class('native', d) and \ bb.data.getVar('PN', d, True) not in \ ("linux-libc-headers-native","quilt-native",
Quoted 8 lines · Michael Smith
Quoted 2 lines · Khem Raj
although it does but make 3.82 is not usable out of box yet. so people might find surprises. 3.81 would be ok.
I could add a
DEFAULT_PREFERENCE= "-1" to the 3.82 recipe and mention the packages that don't build with it in a comment. Do you remember which packages fail to build with 3.82?
well, busybox, glibc, gst* have failed and were fixed. In general I have hit:
- mixing 'rule %rule:', newer GNU make refuses to match that (busybox, glibc)
- spaces vs. tab. e.g. using 8 spaces instead of a tab (gst)
Quoted 10 lines · Holger Freyther
Quoted 3 lines · Michael Smith
I could add a
DEFAULT_PREFERENCE= "-1" to the 3.82 recipe and mention the packages that don't build with it in a comment. Do you remember which packages fail to build with 3.82?well, busybox, glibc, gst* have failed and were fixed. In general I have hit:
- mixing 'rule %rule:', newer GNU make refuses to match that (busybox, glibc)
- spaces vs. tab. e.g. using 8 spaces instead of a tab (gst)
Maybe it'd be easier to fix the affected packages, then. Would you be opposed to make-native being a default dep? The next testing cycle would probably spot any remaining fallout from 3.82.
Mike
Quoted 13 lines · Michael Smith
Quoted 12 lines · Holger Freyther
Quoted 6 lines · Michael Smith
I could add a
DEFAULT_PREFERENCE= "-1" to the 3.82 recipe and mention the
packages that don't build with it in a comment. Do you remember which packages
fail to build with 3.82?well, busybox, glibc, gst* have failed and were fixed. In general I have hit:
Quoted 8 lines
Quoted 3 lines
- mixing 'rule %rule:', newer GNU make refuses to match that (busybox, glibc)
- spaces vs. tab. e.g. using 8 spaces instead of a tab (gst)Maybe it'd be easier to fix the affected packages, then. Would you be opposed to make-native being a default dep? The next testing cycle would probably spot any remaining fallout from 3.82.
make is a core component. We have enough fallout to fix which is not make related. this will add to it. and make 3.82 native recipe is there. I am just not ready to make it default yet.
Quoted 7 lines
Mike
_______________________________________________
Openembedded-devel mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
Signature · 1 line
-Khem
PATCH 1/4 make 3.82: set DEFAULT_PREFERENCE = "-1"
This version has introduced several backwards incompatibilities and some packages are known not to build with 3.82. In particular:
- glibc-initial fails in do_install when prefix="".
- eglibc manual/Makefile has not yet been patched (see 3a06f50f)
--- recipes/make/make_3.82.bb | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
recipes/make/make_3.82.bb +3 -0
@@ -1,5 +1,8 @@ require make.inc +# Stock eglibc fails to build with make 3.82. +DEFAULT_PREFERENCE = "-1" + BBCLASSEXTEND = "native" SRC_URI[md5sum] = "1a11100f3c63fcf5753818e59d63088f" SRC_URI[sha256sum] = "e2c1a73f179c40c71e2fe8abf8a8a0688b8499538512984da4a76958d0402966"
Signature · 1 line
1.7.0.4
PATCH 2/4 base.bbclass: add make-native to default dependencies
This should work around package incompatibilities with whatever version of make is installed on the build system.
--- classes/base.bbclass | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-)
classes/base.bbclass +2 -1
@@ -71,7 +71,8 @@ def base_deps(d): if not bb.data.getVar('INHIBIT_DEFAULT_DEPS', d): if (bb.data.getVar('HOST_SYS', d, 1) != bb.data.getVar('BUILD_SYS', d, 1)): - deps += " virtual/${TARGET_PREFIX}gcc virtual/libc " + deps += (" virtual/${TARGET_PREFIX}gcc virtual/libc " + "make-native") elif bb.data.inherits_class('native', d) and \ bb.data.getVar('PN', d, True) not in \ ("linux-libc-headers-native", "quilt-native",
Signature · 1 line
1.7.0.4
PATCH 3/4 glibc, glibc-initial: depend on make-native
In case the version of make installed on the build host is incompatible, we'll use make-native.
--- recipes/glibc/glibc-initial.inc | 2 +- recipes/glibc/glibc.inc | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-)
recipes/glibc/glibc-initial.inc +1 -1
@@ -1,5 +1,5 @@ SECTION = "libs" -DEPENDS = "virtual/${TARGET_PREFIX}gcc-initial linux-libc-headers" +DEPENDS = "virtual/${TARGET_PREFIX}gcc-initial linux-libc-headers make-native" PROVIDES = "virtual/${TARGET_PREFIX}libc-initial" PACKAGES = ""
recipes/glibc/glibc.inc +3 -1
@@ -4,7 +4,9 @@ SECTION = "libs" PRIORITY = "required" LICENSE = "LGPL" # nptl needs unwind support in gcc, which can't be built without glibc. -DEPENDS = "virtual/${TARGET_PREFIX}gcc-intermediate linux-libc-headers" +DEPENDS = "virtual/${TARGET_PREFIX}gcc-intermediate \ + linux-libc-headers \ + make-native" #this leads to circular deps, so lets not add it yet #RDEPENDS_ldd += " bash" PROVIDES = "virtual/libc virtual/${TARGET_PREFIX}libc-for-gcc"
Signature · 1 line
1.7.0.4
PATCH 4/4 eglibc, eglibc-initial: depend on make-native
In case the version of make installed on the build host is incompatible, we'll use make-native.
--- recipes/eglibc/eglibc-initial.inc | 2 +- recipes/eglibc/eglibc.inc | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-)
recipes/eglibc/eglibc-initial.inc +1 -1
@@ -1,5 +1,5 @@ SECTION = "libs" -DEPENDS = "linux-libc-headers virtual/${TARGET_PREFIX}gcc-initial" +DEPENDS = "linux-libc-headers virtual/${TARGET_PREFIX}gcc-initial make-native" PROVIDES = "virtual/${TARGET_PREFIX}libc-initial" PACKAGES = ""
recipes/eglibc/eglibc.inc +3 -1
@@ -5,7 +5,9 @@ PRIORITY = "required" LICENSE = "LGPL" INC_PR = "r11" # nptl needs unwind support in gcc, which can't be built without glibc. -DEPENDS = "virtual/${TARGET_PREFIX}gcc-intermediate linux-libc-headers" +DEPENDS = "virtual/${TARGET_PREFIX}gcc-intermediate \ + linux-libc-headers \ + make-native" #this leads to circular deps, so lets not add it yet #RDEPENDS_ldd += " bash" # nptl needs libgcc but dlopens it, so our shlibs code doesn't detect this
Signature · 1 line
1.7.0.4
Am Freitag, den 24.09.2010, 15:39 -0400 schrieb Michael Smith:
Quoted 13 lines
This should work around package incompatibilities with whatever version of make is installed on the build system.
Signed-off-by Michael Smith <[email protected]>--- classes/base.bbclass | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-)classes/base.bbclass +0
-0
classes/base.bbclass def base_deps(d): · line 75
Quoted 6 lines
classes/base.bbclass +2
-1@@ -71,7 +71,8 @@ def base_deps(d): if not bb.data.getVar('INHIBIT_DEFAULT_DEPS', d): if (bb.data.getVar('HOST_SYS', d, 1) != bb.data.getVar('BUILD_SYS', d, 1)): - deps += " virtual/${TARGET_PREFIX}gcc virtual/libc " + deps += (" virtual/${TARGET_PREFIX}gcc virtual/libc " + "make-native")
Are the parentheses »()« needed? Is it to prevent using »\« for the line break?
classes/base.bbclass def base_deps(d): · line 78
Quoted 3 lines
classes/base.bbclass +0
-0elif bb.data.inherits_class('native', d) and \ bb.data.getVar('PN', d, True) not in \ ("linux-libc-headers-native", "quilt-native",
Thanks,
Paul
Quoted 10 lines · Paul Menzel
Am Freitag, den 24.09.2010, 15:39 -0400 schrieb Michael Smith:
classes/base.bbclass def base_deps(d): · line 75
Quoted 6 lines
This should work around package incompatibilities with whatever version of make is installed on the build system.
classes/base.bbclass +2
-1- deps += " virtual/${TARGET_PREFIX}gcc virtual/libc " + deps += (" virtual/${TARGET_PREFIX}gcc virtual/libc " + "make-native")Are the parentheses »()« needed? Is it to prevent using »\« for the line break?
Yes, that's correct.
Mike