P3
HI
I currently work for Sun Web Server team and we have a bug that we believe could be in the javac module in Niagara hardware. Can you kindly take a look at this and let us know, your thoughts on this. thanks.
Web Server 6.1 sp5 64-bit is going to be released as part of Spec Web'2005 benchmark tests to show case the performance of Niagara hardware. The spec web'2005 benchmark tests include running the web server under extremely high load so as to serve dynamic web pages like jsp/servlet. (just for the record, these dynamic jsp pages will be converted into plain java file and web server servlet container will be using javac to compile it during runtime. )
PAE team , headed by Pallab Bhattacharya was leading the performance work for Web Server on Niagara hardware. They were using JDK 6 as their default jdk for their performance metrics until the Niagara specific performance enhancements has been back ported to jdk 5.0u6. This team has reported that as soon as they use JDK 5.0u6 (which probably contains Niagara specific enhancements) is throwing random java compilation error during the Spec Web'2005.
They have also observed that once the code gets compiled properly using either jdk 6 or jdk 5.0u5 then there is no problem at all. It seems that there is some issues with javac reporting erroneous compiler error if run under any stress.
Well, now I am able to reproduce this as well. This is what I did to reproduce this
Web Server internally has a regression tool called GAT which can hit the web server with about 100-200 simple jsp pages using around 50 threads. With this setup, I am able to reproduce the random errorneous compiler error as well. This same error almost never happens on a SPARC hardware and consistently happens under Niagara hardware. This issue can't be seen if use only 1 thread. It needs to be more than 1 thread (say 50) to reproduce this issue.
I understand that this description might not be good enough for your team to go on. So, I am currently trying to put together a simple test case as early as possible. I will update the bug report as soon as I manage to put together some.
Looking for your team's comments/thoughts
thanks in advance
sriram.
Now, these tests were running fine until the PAE team cho
I currently work for Sun Web Server team and we have a bug that we believe could be in the javac module in Niagara hardware. Can you kindly take a look at this and let us know, your thoughts on this. thanks.
Web Server 6.1 sp5 64-bit is going to be released as part of Spec Web'2005 benchmark tests to show case the performance of Niagara hardware. The spec web'2005 benchmark tests include running the web server under extremely high load so as to serve dynamic web pages like jsp/servlet. (just for the record, these dynamic jsp pages will be converted into plain java file and web server servlet container will be using javac to compile it during runtime. )
PAE team , headed by Pallab Bhattacharya was leading the performance work for Web Server on Niagara hardware. They were using JDK 6 as their default jdk for their performance metrics until the Niagara specific performance enhancements has been back ported to jdk 5.0u6. This team has reported that as soon as they use JDK 5.0u6 (which probably contains Niagara specific enhancements) is throwing random java compilation error during the Spec Web'2005.
They have also observed that once the code gets compiled properly using either jdk 6 or jdk 5.0u5 then there is no problem at all. It seems that there is some issues with javac reporting erroneous compiler error if run under any stress.
Well, now I am able to reproduce this as well. This is what I did to reproduce this
Web Server internally has a regression tool called GAT which can hit the web server with about 100-200 simple jsp pages using around 50 threads. With this setup, I am able to reproduce the random errorneous compiler error as well. This same error almost never happens on a SPARC hardware and consistently happens under Niagara hardware. This issue can't be seen if use only 1 thread. It needs to be more than 1 thread (say 50) to reproduce this issue.
I understand that this description might not be good enough for your team to go on. So, I am currently trying to put together a simple test case as early as possible. I will update the bug report as soon as I manage to put together some.
Looking for your team's comments/thoughts
thanks in advance
sriram.
Now, these tests were running fine until the PAE team cho