Skip to contentSkip to conversation

mesa: Fix build with upcoming LLVM 17

PATCH Superseded Khem Raj 4 participants 10 messages Jun 24, 2023 – Jul 5, 2023

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Khem Raj

author
#0 lore raw

PATCH 1/3 mesa: Fix build with upcoming LLVM 17

---
 .../0001-gallium-Fix-build-with-llvm-17.patch | 34 +++++++++++++++++++
 meta/recipes-graphics/mesa/mesa.inc           |  6 ++++
 2 files changed, 40 insertions(+)
 create mode 100644 meta/recipes-graphics/mesa/files/0001-gallium-Fix-build-with-llvm-17.patch
meta/recipes-graphics/mesa/files/0001-gallium-Fix-build-with-llvm-17.patch new file +34 -0
new file mode 100644
index 00000000000..3631a918b0c
@@ -0,0 +1,34 @@
+From 865762e0a767a121206d818bdd58301afbf30104 Mon Sep 17 00:00:00 2001
+From: Khem Raj <[email protected]>
+Date: Fri, 23 Jun 2023 01:20:38 -0700
+Subject: [PATCH] gallium: Fix build with llvm 17
+
+These headers are not available for C files in llvm 17+
+and they seem to be not needed to compile after all with llvm 17
+so add conditions to exclude them for llvm >= 17
+
+Upstream-Status: Submitted [https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23827]
+Signed-off-by: Khem Raj <[email protected]>
+---
+ src/gallium/auxiliary/gallivm/lp_bld_init.c | 4 +++-
+ 1 file changed, 3 insertions(+), 1 deletion(-)
+
+diff --git a/src/gallium/auxiliary/gallivm/lp_bld_init.c b/src/gallium/auxiliary/gallivm/lp_bld_init.c
+index 24d0823..3d4573e 100644
+--- a/src/gallium/auxiliary/gallivm/lp_bld_init.c
++++ b/src/gallium/auxiliary/gallivm/lp_bld_init.c
+@@ -42,8 +42,10 @@
+ 
+ #include <llvm/Config/llvm-config.h>
+ #include <llvm-c/Analysis.h>
++#if LLVM_VERSION_MAJOR < 17
+ #include <llvm-c/Transforms/Scalar.h>
+-#if LLVM_VERSION_MAJOR >= 7
++#endif
++#if LLVM_VERSION_MAJOR >= 7 && LLVM_VERSION_MAJOR < 17
+ #include <llvm-c/Transforms/Utils.h>
+ #endif
+ #include <llvm-c/BitWriter.h>
+-- 
+2.41.0
+
meta/recipes-graphics/mesa/mesa.inc +6 -0
index ac42a8dd4f3..c988c77115b 100644
@@ -17,6 +17,7 @@ PE = "2"
 SRC_URI = "https://mesa.freedesktop.org/archive/mesa-${PV}.tar.xz \
            file://0001-meson.build-check-for-all-linux-host_os-combinations.patch \
            file://0001-meson-misdetects-64bit-atomics-on-mips-clang.patch \
+           file://0001-gallium-Fix-build-with-llvm-17.patch \
            "
 
 SRC_URI[sha256sum] = "a2679031ed5b73b29c4f042ac64d96f83b0cfe4858617de32e2efc196c653a40"
@@ -74,6 +75,11 @@ EXTRA_OEMESON = " \
 def strip_comma(s):
     return s.strip(',')
 
+# llvm-config should be the one from llvm-native and not from build host ( if llvm is installed on host )
+do_write_config:append() {
+    sed -i -e '/\[binaries\]/a\' -e "llvm-config = '${STAGING_BINDIR_NATIVE}/llvm-config'" ${WORKDIR}/meson.native
+}
+
 PACKAGECONFIG = " \
 	gallium \
 	${@bb.utils.filter('DISTRO_FEATURES', 'x11 vulkan wayland', d)} \
Signature · 1 line
2.41.0
author
#1 lore raw

PATCH 2/3 libxcrypt: Limit to enabling strong hash algos only

This ensures that weak algorithms are not included, which should improve the defauls to be more secure

---
 meta/recipes-core/libxcrypt/libxcrypt.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes-core/libxcrypt/libxcrypt.inc +1 -1
index 997e83fb004..db43a34b11d 100644
@@ -29,7 +29,7 @@ BUILD_CPPFLAGS = "-I${STAGING_INCDIR_NATIVE}"
 TARGET_CPPFLAGS = "-I${STAGING_DIR_TARGET}${includedir} -Wno-error"
 CPPFLAGS:append:class-nativesdk = " -Wno-error"
 
-API = "--disable-obsolete-api"
+API = "--disable-obsolete-api --enable-hashes=strong"
 EXTRA_OECONF += "${API}"
 
 BBCLASSEXTEND = "native nativesdk"
Signature · 1 line
2.41.0
#2 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

Hello Khem,

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

Quoted 34 lines · Khem Raj

This ensures that weak algorithms are not included, which should improve the defauls to be more secure

---
 meta/recipes-core/libxcrypt/libxcrypt.inc | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes-core/libxcrypt/libxcrypt.inc +1 -1
index 997e83fb004..db43a34b11d 100644
@@ -29,7 +29,7 @@ BUILD_CPPFLAGS = "-I${STAGING_INCDIR_NATIVE}"
 TARGET_CPPFLAGS = "-I${STAGING_DIR_TARGET}${includedir} -Wno-error"
 CPPFLAGS:append:class-nativesdk = " -Wno-error"
 
-API = "--disable-obsolete-api"
+API = "--disable-obsolete-api --enable-hashes=strong"
 EXTRA_OECONF += "${API}"
 
 BBCLASSEXTEND = "native nativesdk"
Signature · 11 lines
2.41.0



-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#183356): https://lists.openembedded.org/g/openembedded-core/message/183356
Mute This Topic: https://lists.openembedded.org/mt/99749300/3617179
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
Signature · 3 lines
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
#3 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

Quoted 8 lines · Alexandre Belloni via lists.openembedded.org

Hello Khem,

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

I would say to send a patch upstream to make those tests not depend on the unsafe algorithms existing, but the entire module has been removed in the next release of Python…

Ross

#4 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

I would also suggest that xcrypt itself is changed to default only to strong algorithms, and weak algorithms is an opt-in.

This idea of 'letting users decide' which crypto they want from all-included set is frankly maddening, and I do not understand why it's still happening in 2023.

Alex

Quoted 22 lines · Ross Burton
Quoted 8 lines · Alexandre Belloni via lists.openembedded.org

Hello Khem,

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

I would say to send a patch upstream to make those tests not depend on the unsafe algorithms existing, but the entire module has been removed in the next release of Python…

Ross
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#183825): https://lists.openembedded.org/g/openembedded-core/message/183825
Mute This Topic: https://lists.openembedded.org/mt/99749300/1686489
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

author
#5 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

In reply to Ross Burton
Quoted 5 lines · Ross Burton
Quoted 3 lines · Alexandre Belloni via lists.openembedded.org

Hello Khem,

Quoted 8 lines
Quoted 5 lines

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

I would say to send a patch upstream to make those tests not depend on the unsafe algorithms existing, but the entire module has been removed in the next release of Python…

crypt is already deprecated in python 11 and will be removed from python 13. [1] some packages have opened issues to track it e.g. [2]

[1] https://peps.python.org/pep-0594/#crypt
[2] https://github.com/saltstack/salt/issues/64263

This is good but perhaps a bit early for us to act upon it, since this would require packages to migrate away from it to use hashlib or some such equivalent module.

Quoted 1 line

Ross

#6 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

In reply to Alexandre Belloni
Quoted 8 lines · Alexandre Belloni via lists.openembedded.org

Hello Khem,

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

Guessing this is also the cause of the perl failures:

{'perl': ['t/op/crypt']}

and apr-util:

{'apr-util': ['testpass_3_of_6']}

Test 3 of 6 is test_shapass using SHA1.

If we want to do this then some effort will be needed to make all the test suites adapt to what algorithms are actually available.

Ross

#7 lore raw

Re: [OE-core] [PATCH 2/3] libxcrypt: Limit to enabling strong hash algos only

Quoted 12 lines · Ross Burton
Quoted 8 lines · Alexandre Belloni via lists.openembedded.org

Hello Khem,

This causes python ptest failures:

{'python3': ['test_crypt', 'test_methods', 'test_sha2_rounds']}

We should probably disable those

Guessing this is also the cause of the perl failures:

Quoted 8 lines

{'perl': ['t/op/crypt']}

and apr-util:

{'apr-util': ['testpass_3_of_6']}

Test 3 of 6 is test_shapass using SHA1.

Yes it is, I should have added that. I dropped the patch for now.

Quoted 15 lines

If we want to do this then some effort will be needed to make all the test suites adapt to what algorithms are actually available.

Ross

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#183887): https://lists.openembedded.org/g/openembedded-core/message/183887
Mute This Topic: https://lists.openembedded.org/mt/99749300/3617179
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Signature · 3 lines
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
author
#8 lore raw

PATCH 3/3 alsa-lib: Disable old API symbols

In reply to Khem Raj

These symbols are currently generated as undefined in the shared object, which means we really do not need them.

---
 meta/recipes-multimedia/alsa/alsa-lib_1.2.9.bb | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
meta/recipes-multimedia/alsa/alsa-lib_1.2.9.bb +1 -1
index a8bf09904d9..d482e27683b 100644
@@ -16,7 +16,7 @@ inherit autotools pkgconfig
 
 EXTRA_OECONF += " \
     ${@bb.utils.contains('TARGET_FPU', 'soft', '--with-softfloat', '', d)} \
-    --disable-python \
+    --disable-python --disable-old-symbols \
 "
 
 PACKAGES =+ "alsa-server alsa-conf libatopology"
Signature · 1 line
2.41.0
#9 lore raw
In reply to Khem Raj

meta/recipes-graphics/mesa/mesa.inc EXTRA_OEMESON = " \ · line 82

Quoted 5 lines · Khem Raj
meta/recipes-graphics/mesa/mesa.inc +5 -0
+# llvm-config should be the one from llvm-native and not from build host ( if llvm is installed on host )
+do_write_config:append() {
+    sed -i -e '/\[binaries\]/a\' -e "llvm-config = '${STAGING_BINDIR_NATIVE}/llvm-config'" ${WORKDIR}/meson.native
+}
+

This change should be in a separate commit, and with a separate explanation please. Is there a mistake in how meson.native is formed, and if so, should it be fixed at the source?

Alex