Re: [10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
I can just adjust that before I push. I'm running it through internal testing now. Regarding quoting, make doesn't really do quotes. When you see quoted strings in make, it's generally done for shell consumption. /Erik On 2018-01-17 13:55, David Holmes wrote: Hi Volker, Changes seem okay to me too. spec.gmk.in: ! # Only export VENDOR_URL, VENDOR_URL_BUG and VENDOR_VM_URL_BUG tot he build if they Typo: tot he -> to the I'm also surprised this doesn't need any quoting: ifneq ($(COMPANY_NAME), N/A) Thanks, David On 18/01/2018 2:03 AM, Volker Simonis wrote: Hi, can I please have a review and sponsor for this change which finally exposes the various "vendor*" properties: java.vendor java.vm.vendor java.vendor.url java.vendor.url.bug as configure arguments: http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 https://bugs.openjdk.java.net/browse/JDK-8189761 With this change, "java.vendor" and "java.vm.vendor" still default to "Oracle Corporation" which is defined in System.c (for "java.vendor") and in vm_version.cpp (for "java.vm.version") unless the new "--with-vendor-name" option introduced by JDK-8193764 will be used in the configure step. If the "--with-vendor-name" option will be used, its value will now, with this change, also be assigned to both, the "java.vendor" and "java.vm.vendor" properties. I don't think that we need a separate configure option for "java.vm.vendor" because if somebody is building the OpenJDK with a different VM, he will own the source code of that VM anyway and can easily set "java.vm.vendor" in his code. For "java.vendor.url" and "java.vendor.url.bug" I've introduced the two new configure options "--with-vendor-url" and "--with-vendor-bug-url" which should be self explanatory. If they are not set, the old default values will be used. In the case of a VM crash, the HotSpot displays a second, different bug URL which is currently only configured in the arguments.cpp file. I've exposed this URL as well to configure with the new "--with-vendor-vm-bug-url" configure option. Again, if this option will not use, the VM will fall back to the old default value. Notice that this patch also fixes a bug introduced by "8193764: Cannot set COMPANY_NAME when configuring a build" because of which the jtreg test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the testee was configured without "--with-vendor-name". The problem is that the code introduced by 8193764 unconditionally sets COMPANY_NAME to the empty string, if no "--with-vendor-name" option was given. This overrides the default setting of COMPANY_NAME=N/A from $AUTOCONF_DIR/version-numbers. I want to bring this into jdk10 if possible. I need a sponsor because this change requires the regeneration of "generated-configure.sh" (which I've included in the webrev for your convenience only) and because it touches a HotSpot file and external contributors still can't push such changes :( Thank you and best regards, Volker
Re: [10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
David Holmesschrieb am Mi. 17. Jan. 2018 um 22:55: > Hi Volker, > > Changes seem okay to me too. > Thanks David! > spec.gmk.in: > > ! # Only export VENDOR_URL, VENDOR_URL_BUG and VENDOR_VM_URL_BUG tot he > build if they > > Typo: tot he -> to the > Will fix tomorrow morning when I’m back in the office! > I'm also surprised this doesn't need any quoting: > >ifneq ($(COMPANY_NAME), N/A) > Me too, but I can assure you that it doesn’t work if quoted. I’ve tried it :) > Thanks, > David > > On 18/01/2018 2:03 AM, Volker Simonis wrote: > > Hi, > > > > can I please have a review and sponsor for this change which finally > > exposes the various "vendor*" properties: > > > > java.vendor > > java.vm.vendor > > java.vendor.url > > java.vendor.url.bug > > > > as configure arguments: > > > > http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 > > https://bugs.openjdk.java.net/browse/JDK-8189761 > > > > With this change, "java.vendor" and "java.vm.vendor" still default to > > "Oracle Corporation" which is defined in System.c (for "java.vendor") > > and in vm_version.cpp (for "java.vm.version") unless the new > > "--with-vendor-name" option introduced by JDK-8193764 will be used in > > the configure step. > > > > If the "--with-vendor-name" option will be used, its value will now, > > with this change, also be assigned to both, the "java.vendor" and > > "java.vm.vendor" properties. I don't think that we need a separate > > configure option for "java.vm.vendor" because if somebody is building > > the OpenJDK with a different VM, he will own the source code of that > > VM anyway and can easily set "java.vm.vendor" in his code. > > > > For "java.vendor.url" and "java.vendor.url.bug" I've introduced the > > two new configure options "--with-vendor-url" and > > "--with-vendor-bug-url" which should be self explanatory. If they are > > not set, the old default values will be used. > > > > In the case of a VM crash, the HotSpot displays a second, different > > bug URL which is currently only configured in the arguments.cpp file. > > I've exposed this URL as well to configure with the new > > "--with-vendor-vm-bug-url" configure option. Again, if this option > > will not use, the VM will fall back to the old default value. > > > > Notice that this patch also fixes a bug introduced by "8193764: Cannot > > set COMPANY_NAME when configuring a build" because of which the jtreg > > test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the > > testee was configured without "--with-vendor-name". The problem is > > that the code introduced by 8193764 unconditionally sets COMPANY_NAME > > to the empty string, if no "--with-vendor-name" option was given. This > > overrides the default setting of COMPANY_NAME=N/A from > > $AUTOCONF_DIR/version-numbers. > > > > I want to bring this into jdk10 if possible. > > > > I need a sponsor because this change requires the regeneration of > > "generated-configure.sh" (which I've included in the webrev for your > > convenience only) and because it touches a HotSpot file and external > > contributors still can't push such changes :( > > > > Thank you and best regards, > > Volker > > >
Re: [10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
Hi Volker, Changes seem okay to me too. spec.gmk.in: ! # Only export VENDOR_URL, VENDOR_URL_BUG and VENDOR_VM_URL_BUG tot he build if they Typo: tot he -> to the I'm also surprised this doesn't need any quoting: ifneq ($(COMPANY_NAME), N/A) Thanks, David On 18/01/2018 2:03 AM, Volker Simonis wrote: Hi, can I please have a review and sponsor for this change which finally exposes the various "vendor*" properties: java.vendor java.vm.vendor java.vendor.url java.vendor.url.bug as configure arguments: http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 https://bugs.openjdk.java.net/browse/JDK-8189761 With this change, "java.vendor" and "java.vm.vendor" still default to "Oracle Corporation" which is defined in System.c (for "java.vendor") and in vm_version.cpp (for "java.vm.version") unless the new "--with-vendor-name" option introduced by JDK-8193764 will be used in the configure step. If the "--with-vendor-name" option will be used, its value will now, with this change, also be assigned to both, the "java.vendor" and "java.vm.vendor" properties. I don't think that we need a separate configure option for "java.vm.vendor" because if somebody is building the OpenJDK with a different VM, he will own the source code of that VM anyway and can easily set "java.vm.vendor" in his code. For "java.vendor.url" and "java.vendor.url.bug" I've introduced the two new configure options "--with-vendor-url" and "--with-vendor-bug-url" which should be self explanatory. If they are not set, the old default values will be used. In the case of a VM crash, the HotSpot displays a second, different bug URL which is currently only configured in the arguments.cpp file. I've exposed this URL as well to configure with the new "--with-vendor-vm-bug-url" configure option. Again, if this option will not use, the VM will fall back to the old default value. Notice that this patch also fixes a bug introduced by "8193764: Cannot set COMPANY_NAME when configuring a build" because of which the jtreg test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the testee was configured without "--with-vendor-name". The problem is that the code introduced by 8193764 unconditionally sets COMPANY_NAME to the empty string, if no "--with-vendor-name" option was given. This overrides the default setting of COMPANY_NAME=N/A from $AUTOCONF_DIR/version-numbers. I want to bring this into jdk10 if possible. I need a sponsor because this change requires the regeneration of "generated-configure.sh" (which I've included in the webrev for your convenience only) and because it touches a HotSpot file and external contributors still can't push such changes :( Thank you and best regards, Volker
Re: [10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
Erik Joelssonschrieb am Mi. 17. Jan. 2018 um 20:16: > Looks good to me. I can sponsor it. Thanks a lot Erik! Regards, Volker > > /Erik > > > On 2018-01-17 08:03, Volker Simonis wrote: > > Hi, > > > > can I please have a review and sponsor for this change which finally > > exposes the various "vendor*" properties: > > > > java.vendor > > java.vm.vendor > > java.vendor.url > > java.vendor.url.bug > > > > as configure arguments: > > > > http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 > > https://bugs.openjdk.java.net/browse/JDK-8189761 > > > > With this change, "java.vendor" and "java.vm.vendor" still default to > > "Oracle Corporation" which is defined in System.c (for "java.vendor") > > and in vm_version.cpp (for "java.vm.version") unless the new > > "--with-vendor-name" option introduced by JDK-8193764 will be used in > > the configure step. > > > > If the "--with-vendor-name" option will be used, its value will now, > > with this change, also be assigned to both, the "java.vendor" and > > "java.vm.vendor" properties. I don't think that we need a separate > > configure option for "java.vm.vendor" because if somebody is building > > the OpenJDK with a different VM, he will own the source code of that > > VM anyway and can easily set "java.vm.vendor" in his code. > > > > For "java.vendor.url" and "java.vendor.url.bug" I've introduced the > > two new configure options "--with-vendor-url" and > > "--with-vendor-bug-url" which should be self explanatory. If they are > > not set, the old default values will be used. > > > > In the case of a VM crash, the HotSpot displays a second, different > > bug URL which is currently only configured in the arguments.cpp file. > > I've exposed this URL as well to configure with the new > > "--with-vendor-vm-bug-url" configure option. Again, if this option > > will not use, the VM will fall back to the old default value. > > > > Notice that this patch also fixes a bug introduced by "8193764: Cannot > > set COMPANY_NAME when configuring a build" because of which the jtreg > > test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the > > testee was configured without "--with-vendor-name". The problem is > > that the code introduced by 8193764 unconditionally sets COMPANY_NAME > > to the empty string, if no "--with-vendor-name" option was given. This > > overrides the default setting of COMPANY_NAME=N/A from > > $AUTOCONF_DIR/version-numbers. > > > > I want to bring this into jdk10 if possible. > > > > I need a sponsor because this change requires the regeneration of > > "generated-configure.sh" (which I've included in the webrev for your > > convenience only) and because it touches a HotSpot file and external > > contributors still can't push such changes :( > > > > Thank you and best regards, > > Volker > >
Re: [10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
Looks good to me. I can sponsor it. /Erik On 2018-01-17 08:03, Volker Simonis wrote: Hi, can I please have a review and sponsor for this change which finally exposes the various "vendor*" properties: java.vendor java.vm.vendor java.vendor.url java.vendor.url.bug as configure arguments: http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 https://bugs.openjdk.java.net/browse/JDK-8189761 With this change, "java.vendor" and "java.vm.vendor" still default to "Oracle Corporation" which is defined in System.c (for "java.vendor") and in vm_version.cpp (for "java.vm.version") unless the new "--with-vendor-name" option introduced by JDK-8193764 will be used in the configure step. If the "--with-vendor-name" option will be used, its value will now, with this change, also be assigned to both, the "java.vendor" and "java.vm.vendor" properties. I don't think that we need a separate configure option for "java.vm.vendor" because if somebody is building the OpenJDK with a different VM, he will own the source code of that VM anyway and can easily set "java.vm.vendor" in his code. For "java.vendor.url" and "java.vendor.url.bug" I've introduced the two new configure options "--with-vendor-url" and "--with-vendor-bug-url" which should be self explanatory. If they are not set, the old default values will be used. In the case of a VM crash, the HotSpot displays a second, different bug URL which is currently only configured in the arguments.cpp file. I've exposed this URL as well to configure with the new "--with-vendor-vm-bug-url" configure option. Again, if this option will not use, the VM will fall back to the old default value. Notice that this patch also fixes a bug introduced by "8193764: Cannot set COMPANY_NAME when configuring a build" because of which the jtreg test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the testee was configured without "--with-vendor-name". The problem is that the code introduced by 8193764 unconditionally sets COMPANY_NAME to the empty string, if no "--with-vendor-name" option was given. This overrides the default setting of COMPANY_NAME=N/A from $AUTOCONF_DIR/version-numbers. I want to bring this into jdk10 if possible. I need a sponsor because this change requires the regeneration of "generated-configure.sh" (which I've included in the webrev for your convenience only) and because it touches a HotSpot file and external contributors still can't push such changes :( Thank you and best regards, Volker
[10] RFR(S): 8189761: COMPANY_NAME, IMPLEMENTOR, BUNDLE_VENDOR, VENDOR, but no configure flag
Hi, can I please have a review and sponsor for this change which finally exposes the various "vendor*" properties: java.vendor java.vm.vendor java.vendor.url java.vendor.url.bug as configure arguments: http://cr.openjdk.java.net/~simonis/webrevs/2018/8189761 https://bugs.openjdk.java.net/browse/JDK-8189761 With this change, "java.vendor" and "java.vm.vendor" still default to "Oracle Corporation" which is defined in System.c (for "java.vendor") and in vm_version.cpp (for "java.vm.version") unless the new "--with-vendor-name" option introduced by JDK-8193764 will be used in the configure step. If the "--with-vendor-name" option will be used, its value will now, with this change, also be assigned to both, the "java.vendor" and "java.vm.vendor" properties. I don't think that we need a separate configure option for "java.vm.vendor" because if somebody is building the OpenJDK with a different VM, he will own the source code of that VM anyway and can easily set "java.vm.vendor" in his code. For "java.vendor.url" and "java.vendor.url.bug" I've introduced the two new configure options "--with-vendor-url" and "--with-vendor-bug-url" which should be self explanatory. If they are not set, the old default values will be used. In the case of a VM crash, the HotSpot displays a second, different bug URL which is currently only configured in the arguments.cpp file. I've exposed this URL as well to configure with the new "--with-vendor-vm-bug-url" configure option. Again, if this option will not use, the VM will fall back to the old default value. Notice that this patch also fixes a bug introduced by "8193764: Cannot set COMPANY_NAME when configuring a build" because of which the jtreg test "test/jdk/tools/jlink/ReleaseImplementorTest.java" fails if the testee was configured without "--with-vendor-name". The problem is that the code introduced by 8193764 unconditionally sets COMPANY_NAME to the empty string, if no "--with-vendor-name" option was given. This overrides the default setting of COMPANY_NAME=N/A from $AUTOCONF_DIR/version-numbers. I want to bring this into jdk10 if possible. I need a sponsor because this change requires the regeneration of "generated-configure.sh" (which I've included in the webrev for your convenience only) and because it touches a HotSpot file and external contributors still can't push such changes :( Thank you and best regards, Volker