Skip to contentSkip to conversation

openh264: switch away from github archive

PATCH In review Trevor Woerner 3 participants 5 messages May 10, 2018 – May 16, 2018

lore
Share & tools
lore

Trevor Woerner

author
#0 lore raw

PATCH v2 1/1 openh264: switch away from github archive

Since we know that github archives which are automatically generated have a tendency to change their checksums[1], switch to using a git clone.

[1] http://lists.openembedded.org/pipermail/openembedded-devel/2017-September/114916.html

Signed-off-by Trevor Woerner <[email protected]>

---

Changes since v1:
- replace the frankenstein local variable name OPENHBRANCH with simply BRANCH to hold the name of the branch to use

---
 .../recipes-multimedia/openh264/openh264_1.7.0.bb          | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
meta-multimedia/recipes-multimedia/openh264/openh264_1.7.0.bb +4 -3
index a7aaab3623..40bd493e6d 100644
@@ -11,9 +11,10 @@ LICENSE = "BSD-2-Clause"
 LICENSE_FLAGS = "commercial"
 LIC_FILES_CHKSUM = "file://LICENSE;md5=bb6d3771da6a07d33fd50d4d9aa73bcf"
 
-SRC_URI = "https://github.com/cisco/openh264/archive/v${PV}.tar.gz"
-SRC_URI[md5sum] = "93da4e76cfda7ede8fd2df51b0021efd"
-SRC_URI[sha256sum] = "9c07c38d7de00046c9c52b12c76a2af7648b70d05bd5460c8b67f6895738653f"
+S = "${WORKDIR}/git"
+SRCREV = "a180c9d4d6f1a4830ca9eed9d159d54996bd63cb"
+BRANCH = "openh264v1.7"
+SRC_URI = "git://github.com/cisco/openh264.git;protocol=https;branch=${BRANCH};"
 
 COMPATIBLE_MACHINE_armv7a = "(.*)"
 COMPATIBLE_MACHINE_aarch64 = "(.*)"
Signature · 1 line
2.17.0.rc2.3.gc2a499e6c
#1 lore raw
Quoted 11 lines · Trevor Woerner

Since we know that github archives which are automatically generated have a tendency to change their checksums[1], switch to using a git clone.

[1] http://lists.openembedded.org/pipermail/openembedded-devel/2017-September/114916.html

Signed-off-by Trevor Woerner <[email protected]>

---

Changes since v1:
- replace the frankenstein local variable name OPENHBRANCH with simply BRANCH to hold the name of the branch to use

This fails to build: please see
http://errors.yoctoproject.org/Errors/Details/178572/

Quoted 23 lines
---
 .../recipes-multimedia/openh264/openh264_1.7.0.bb          | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
meta-multimedia/recipes-multimedia/openh264/openh264_1.7.0.bb +4 -3
index a7aaab3623..40bd493e6d 100644
@@ -11,9 +11,10 @@ LICENSE = "BSD-2-Clause"
 LICENSE_FLAGS = "commercial"
 LIC_FILES_CHKSUM = "file://LICENSE;md5=bb6d3771da6a07d33fd50d4d9aa73bcf"
 
-SRC_URI = "https://github.com/cisco/openh264/archive/v${PV}.tar.gz"
-SRC_URI[md5sum] = "93da4e76cfda7ede8fd2df51b0021efd"
-SRC_URI[sha256sum] = "9c07c38d7de00046c9c52b12c76a2af7648b70d05bd5460c8b67f6895738653f"
+S = "${WORKDIR}/git"
+SRCREV = "a180c9d4d6f1a4830ca9eed9d159d54996bd63cb"
+BRANCH = "openh264v1.7"
+SRC_URI = "git://github.com/cisco/openh264.git;protocol=https;branch=${BRANCH};"
 
 COMPATIBLE_MACHINE_armv7a = "(.*)"
 COMPATIBLE_MACHINE_aarch64 = "(.*)"
author
#2 lore raw
Quoted 13 lines · On Tue, May 15, 2018 at 10:45 AM, akuster808 <[email protected]> wrote:
Quoted 1 line · Trevor Woerner

Since we know that github archives which are automatically generated

have a

Quoted 3 lines

tendency to change their checksums[1], switch to using a git clone.

[1] http://lists.openembedded.org/pipermail/openembedded-devel/

2017-September/114916.html

Quoted 4 lines
Signed-off-by Trevor Woerner <[email protected]>

---

Quoted 6 lines
Quoted 2 lines

Changes since v1:
- replace the frankenstein local variable name OPENHBRANCH with simply

BRANCH

Quoted 1 line

to hold the name of the branch to use

This fails to build: please see
http://errors.yoctoproject.org/Errors/Details/178572/

As it turns out, even without this patch, building the existing openh264 recipe in master for qemux86 fails exactly the same way. The really odd thing is, I don't see any indication of this failure in any of the "state of the world" build reports. Do the "state of the world" builds not build against MACHINE="qemux86"? Or maybe openh264 is skipped (perhaps due to LICENSE_FLAGS = "commercial")?

I'm not entirely surprised this patch doesn't introduce a new problem, its only difference against the original is to fetch from git rather that an automatically generated github archive. If the auto-generated archive succeeded but a git clone of the exact same SHA failed, that would be news indeed!

I assume that although my patches aren't introducing a new build failure, I should probably investigate fixing it anyway?

The existing recipe has provisions for i386. If I understand correctly, qemux86 is i586 (?), and meta-intel's intel-core2-32 is i686 (?). Do I duplicate the i386 provisions for i586 and i686? Is there an easier way to specify all 32-bit IA systems? Otherwise I guess another option is to disable 32-bit IA COMPATIBLE_MACHINEs?

#3 lore raw
Quoted 4 lines · Trevor Woerner

The existing recipe has provisions for i386. If I understand correctly, qemux86 is i586 (?), and meta-intel's intel-core2-32 is i686 (?). Do I duplicate the i386 provisions for i586 and i686? Is there an easier way to specify all 32-bit IA systems?

Does it work if you use _x86 instead of _i386?

Thanks,
Anuj

author
#4 lore raw

Hi Anuj,

Quoted 9 lines · Anuj Mittal
Quoted 3 lines · Trevor Woerner

The existing recipe has provisions for i386. If I understand correctly, qemux86 is i586 (?), and meta-intel's intel-core2-32 is i686 (?). Do I duplicate the i386 provisions for i586 and i686? Is there an easier way

to

Quoted 1 line

specify all 32-bit IA systems?

Does it work if you use _x86 instead of _i386?

Yes, that seems to help, thanks!