Folks, https://issues.apache.org/bugzilla/show_bug.cgi?id=46666 highlighted some issues that have now been fixed. In commenting on the fix, Bill said:
<quote> The original was an optimization to not keep setting the timeout to the same value if we haven't changed it. Of course it is broken, but a better patch would be to set the timeout correctly outside the loop, and retain the disableUploadTimeout check. The first request shouldn't care, since we should already have available() > 0 with the JIO connector. </quote> My response was: <quote> I took a look at this. I think you do have to set timeouts inside the loop otherwise the request line read doesn't timeout correctly. I wrote this patch that only sets timeouts when required: http://people.apache.org/~markt/patches/2009-02-12-soTimeout.patch but I'm not sure it is necessary. Is setSoTimeout really that expensive? It takes less than a microsecond on my box. The code would be cleaner if we just took out most of the tests and set the timeout, even if it was already set as required. </quote> So, what do people think. Use the patch above or remove most of the tests and just set the timeout? Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
