Hide Forgot
Always, when any part of java version changed from single digit to two digits, issues with priority rise. It was fixed by comment like eg: #TODO when major version will jump over 9, change 1700 to 170 %define priority 1700%{priority_major_version}%{priority_minor_version} Anyway, this is not enough. during *every* (single to two numbers) release this is violated, and I think the jdks priority length slipped out of synchronization again across various jdks. fix is, to enhance %prep by: prioritylength=`expr length %{priority}` if [ $prioritylength -ne 9 ] ... with some nice outputs and exits eg: --- a/java-1.8.0-openjdk.spec +++ b/java-1.8.0-openjdk.spec @@ -118,7 +118,6 @@ %global aarch64_updatever 40 %global aarch64_buildver b12 %global aarch64_changesetid aarch64-1263 -# priority must be 6 digits in total %global priority 18000%{updatever} %global javaver 1.8.0 @@ -951,6 +950,11 @@ if [ %{include_debug_build} -eq 0 -a %{include_normal_build} -eq 0 ] ; then echo "you have disabled both include_debug_build and include_debug_build. no go." exit 13 fi +prioritylength=`expr length %{priority}` +if [ $prioritylength -ne 9 ] ; then + echo "priority must be 9 digits in total, violated" + exit 14 +fi %setup -q -c -n %{uniquesuffix ""} -T -a 0 %ifarch %{aarch64} pushd jdk8 Which will successfully halt built if priority slips. This bug should be tracking bug for: rhel5 java-1.7.0-openjdk java-1.7.0-oracle java-1.7.0-ibm java-1.6.0-openjdk java-1.6.0-oracle java-1.6.0-ibm java-1.5.0-ibm rhel6 java-1.8.0-openjdk java-1.8.0-oracle java-1.8.0-ibm java-1.7.0-openjdk java-1.7.0-oracle java-1.7.1-ibm java-1.6.0-openjdk java-1.6.0-oracle java-1.6.0-ibm java-1.5.0-ibm rhel7 java-1.8.0-openjdk java-1.8.0-oracle java-1.8.0-ibm java-1.7.0-openjdk java-1.7.0-oracle java-1.7.1-ibm java-1.6.0-openjdk java-1.6.0-oracle java-1.6.0-ibm And fedoras 21,22 java-1.8.0-openjdk And fedoras 20 java-1.7.0-openjdk Few more errors like this, and we will run out of integer space...
The check test is now pushed to f21-f23 openjdk8, with test on *7* As oracles', openjdks' and ibms' jdks each have priority based on different and differently growing numbers (SR+FP for ibm, icedtea foe openjdk<8, u for openjdk>7, u+b for Oracle) The idea of gruping them (so they will be moreover competitive only among themselves) : ibm - length of 7 oracle - lenght of 8 openjdk - length of 9 Thoughts?
Current priorities table: f20 f21 f22 rawhide rhel5 rhel6.5 rhel6.6 rhel7.0 rhel7.1 (openjdk 6) na na na na 5 5 5 5 5 (openjdk 7) 6 na na na 6 6 6 6 6 (opendjk 8) 7 7 7 7 na 6 6 6 6 (oracle 6) na na na na 6 6 6 6 6 (oracle 7) na na na na 6 6 6 6 6 (oracle 8) na na na na na N 6 N 6 (ibm 5) na na na na 6 6 6 na na (ibm 6) na na na na 6 6 6 6 N (ibm 7) na na na na 6 6 6 6 na (ibm 7.1) na na na na N N 6 6 6 Legend: na - package is not live, N - no repositories, number = the current number of digits for priority. Plan - synchronize priorities in all ibm javas to 6 digits, oracle javas to 7 digits, openjdk to 8 digits - alreday done by jvanek on fedora 20 21, 22, 23 for openjdk 8.
hi Jana, I have pushed the check to jdk7: f20 f21 f22 rawhide rhel5 rhel6.5 rhel6.6 rhel7.0 rhel7.1 (openjdk 7) 6 na na na 6 6 6 *7* *7* And will do the same for 8 (opendjk 8) 7 7 7 7 na 6 6 *7* *7* Please, add the checks to to proprietary jdks Well, and keep them all on 6...
also I have pushed the check to jdk7: rhel6.7 rhel7.2 6 *7* And will do the same for 8 6 *7*
Packages with proprietary Oracle jdk have all priority 6 digits long: -jdk 8 on rhel 6.6,6.7,7.1,7.2 -jdk 7 on rhel 5.11,6.6,6.7,7.1,7.2 -jdk 6 on rhel 5.11,6.6,6.7,7.1,7.2
(In reply to jfabriko from comment #7) > Packages with proprietary Oracle jdk have all priority 6 digits long: > -jdk 8 on rhel 6.6,6.7,7.1,7.2 > -jdk 7 on rhel 5.11,6.6,6.7,7.1,7.2 > -jdk 6 on rhel 5.11,6.6,6.7,7.1,7.2 +the priority length check included
(In reply to jiri vanek from comment #6) > also > > > I have pushed the check to jdk7: > rhel6.7 rhel7.2 > 6 *7* > > And will do the same for 8 > 6 *7* And will do the same for 8 6.6 6.7 7.1 7.2 7 *7*
For openjdk11, there will be need of bump to 8, otherwise 11 will not beat 18, which was used for 1.8.0