I have a 2 year old Mac Book Pro, running 10.7. I recently updated and installed Java 7 (from Oracle). /usr/libexec/java_home prints: /Library/Java/JavaVirtualMachines/jdk1.7.0_21.jdk/Contents/Home and I've set JAVA_HOME to that. With r2644, and a clean checkout (removing even ignored files), "ant dist" works fine and the output looks normal. Then, running 'ant test' gets me 8 failures and 13 errors out of 291 tests. So, am I using a reasonable jdk version? are these test faailures of concern (they seem to be) does this work ok for others? Thanks, Greg
Hi Greg, some tests require input files which are not stored in svn. ant test doesn't download them automatically, you have to execute ant obtain-test-input-files first. @Steve: I also run into this problem again and again. Would it be a problem to change build.xml to do the download? Gerd Greg Troxel wrote
With r2644, and a clean checkout (removing even ignored files), "ant dist" works fine and the output looks normal.
Then, running 'ant test' gets me 8 failures and 13 errors out of 291 tests.
-- View this message in context: http://gis.19327.n5.nabble.com/tests-passing-on-trunk-tp5765303p5765318.html Sent from the Mkgmap Development mailing list archive at Nabble.com.
OK that is done. We do appear to have a couple of actual test failures though. There is also a second issue which is preventing the latest build from appearing on the download page. One of the tests in StyleImpl is failing randomly when run on the build machine. I'll have to look into that. ..Steve
On 14/06/13 10:47, Steve Ratcliffe wrote:
We do appear to have a couple of actual test failures though.
What was happening here is that with a clean checkout the test fails, but if any file has been modified it passes. This is because of the version number is in the copyright strings and it varies by having an appended 'M' if the working copy is modified. I've relaxed the strict test for the section size to allow for a couple of bytes of variation so this is OK now. The other failure that prevents the builds going onto the download page is still present. Looking into that now. ..Steve
Hi Steve, don't know if this is related to the test issues: recently I noticed small differences between the img file produced with JRE 1.6 and 1.7 AFAIK the file sizes are eqaul, but the order of a few elements in the RGN file was different. I did not invest much time in that after I found out that I used two different configurations in the test. Gerd Steve Ratcliffe wrote
On 14/06/13 10:47, Steve Ratcliffe wrote:
We do appear to have a couple of actual test failures though.
What was happening here is that with a clean checkout the test fails, but if any file has been modified it passes. This is because of the version number is in the copyright strings and it varies by having an appended 'M' if the working copy is modified.
I've relaxed the strict test for the section size to allow for a couple of bytes of variation so this is OK now.
The other failure that prevents the builds going onto the download page is still present. Looking into that now.
..Steve _______________________________________________ mkgmap-dev mailing list
mkgmap-dev@.org
-- View this message in context: http://gis.19327.n5.nabble.com/tests-passing-on-trunk-tp5765303p5765403.html Sent from the Mkgmap Development mailing list archive at Nabble.com.
Hi Gerd
don't know if this is related to the test issues: recently I noticed small differences between the img file produced with JRE 1.6 and 1.7
That is intesting, but I don't think it is the problem here. I've found that the test that fails is StyleImplTest.testGetInfo() It gets the value 1 for version instead of the value 2.2 which is the version number in teststyles/simple. I thought it might be a difference in environment when called from cron, but sometimes it works and sometimes it doesn't even from cron. At the moment it is building and testing fine, so I have no idea what is going on. I've given up for now. Will just have to keep watch on the nightly builds. ..Steve
Hi Steve, Steve Ratcliffe wrote
It gets the value 1 for version instead of the value 2.2 which is the version number in teststyles/simple.
I saw this problem on my netbook (Win 7, JDK 1.7) as well for quite a while, and I still see it with r2646. I've never seen the problem on my desktop machine running JDK 1.6. FindBugs complains because the field style in is declared static, but modified: style = new StyleImpl(STYLE_LOC, "simple"); could that be the reason? Gerd -- View this message in context: http://gis.19327.n5.nabble.com/tests-passing-on-trunk-tp5765303p5765443.html Sent from the Mkgmap Development mailing list archive at Nabble.com.
Hi Gerd
FindBugs complains because the field style in is declared static, but modified: style = new StyleImpl(STYLE_LOC, "simple"); could that be the reason?
Ahh, yes that is probably it. It is down to a change in Java 7 which is why it has just started happening. http://intellijava.blogspot.co.uk/2012/05/junit-and-java-7.html Thanks, I will fix it. ..Steve
GerdP <gpetermann_muenchen@hotmail.com> writes:
some tests require input files which are not stored in svn. ant test doesn't download them automatically, you have to execute ant obtain-test-input-files first. @Steve: I also run into this problem again and again. Would it be a problem to change build.xml to do the download?
Thanks to you and Steve - I updated, redid 'ant build' and 'ant test' and got a clean test run with zero failures.
participants (3)
-
GerdP -
Greg Troxel -
Steve Ratcliffe