User Tools

Site Tools


build_changes

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
build_changes [2017/10/08 16:05]
Mats Wichmann
build_changes [2017/10/31 19:19] (current)
Mats Wichmann
Line 10: Line 10:
   * Unit-test data collection:   * Unit-test data collection:
     * Insufficient collection of problems: right now it looks like if a test dies with a core dump, that's not recorded in the xml files, and the Jenkins publisher doesn'​t see them, so you get cute things like the overview page saying no failures, while the build log has very noisy statements about a failure. [[https://​jira.iotivity.org/​browse/​IOT-2510|IOT-2510]] The log collection logic ("​publisher"​) is at the end of this page.     * Insufficient collection of problems: right now it looks like if a test dies with a core dump, that's not recorded in the xml files, and the Jenkins publisher doesn'​t see them, so you get cute things like the overview page saying no failures, while the build log has very noisy statements about a failure. [[https://​jira.iotivity.org/​browse/​IOT-2510|IOT-2510]] The log collection logic ("​publisher"​) is at the end of this page.
-    * unit test results from other builders than the one named unit-tests are not collated overall. ​ Is this a problem? +    * unit test results from other builders than the one named unit-tests are not collated overall.  ​Windows and Tizen also run some unit tests. ​Is this a problem? ​ These other platforms may have their own logic for deciding unit-test failure. 
-    * ''​auto_build unit_tests'' ​perform ​back to back non-secure and secure tests. the file names are the same, so the publisher collects only the secure set if both finish running - the non-secure ones are overwritten. ​ That may not normally be a problem, but there could be some differences.+    * ''​auto_build.py unit_tests'' ​performs ​back to back non-secure and secure tests. the generated logfile ​names are the same, so the publisher collects only the secure set if both finish running - the non-secure ones are overwritten. ​ That may not normally be a problem, but there could be some differences.
   * Support selective unit testing (a Jira ticket is about to be filed on this). ​ Affects developers, not Jenkins use.   * Support selective unit testing (a Jira ticket is about to be filed on this). ​ Affects developers, not Jenkins use.
-  * Make whether or not to run tests under valgrid selectable (currently the logic is "if linux, run under valgrind"​)+  * Make whether or not to run tests under valgrid selectable (currently the logic is "if linux, run under valgrind"​).  For example, some developers were building iotivity on Raspberry Pi running Raspbian, which passes the "is it Linux" test, but the platform is way too small to support the valgrind runs.
  
 === Jenins xunit Publisher === === Jenins xunit Publisher ===
build_changes.txt · Last modified: 2017/10/31 19:19 by Mats Wichmann