[jira] [Commented] (TS-1103) Traffic Server ESI plugin issues
[ https://issues.apache.org/jira/browse/TS-1103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13248517#comment-13248517 ] Zhao Yongming commented on TS-1103: --- hmm, I think the codes may need more cleanup, so far we have now a working base. Traffic Server ESI plugin issues Key: TS-1103 URL: https://issues.apache.org/jira/browse/TS-1103 Project: Traffic Server Issue Type: Bug Components: Plugins Affects Versions: sometime Environment: Newest trunk. Reporter: Kevin Fox Assignee: Zhao Yongming Fix For: 3.1.4 Attachments: esi.patch, gzip.patch Patch to fix: * Makefile fix to add missing files. * Change return code checking to match whats trunk trafficserver. * Include missing header files. * Fix c++ namespace issues. * Work around strange name mangling/linking issue. * Force the assumption that the cached data is RAW_ESI, not PACKED_ESI. Things wouldn't work without it. * Comment out a block of code that looked to be incorrectly handling EOF. After this, simply loading the plugin and setting response header X-ESI in apache httpd seems to work. A few further bugs I have bumped into that aren't addressed in this patch: * It doesn't seem to parse gzip like it looks like it should. To work around, I had to disable it in apache httpd with RewriteRule . - [E=no-gzip:1] * If the client requests gzip, the ESI processor will gzip the result. It works in firefox but is invalid in chrome. Pulling a dump with curl and running it through gzip --list shows it has the correct uncompressed size and compressed size. using zcat shows the correct data but has the warning: invalid compressed data--length error. As far as I read the gzip spec though, the raw binary file looks valid to me. Not sure what this is. This can probably be simply disabled for now though. * esi:include is slightly broken. You get all the data back properly but sometimes the headers are sent prematurely with a Content-Length of 2**31-1. This causes clients to timeout and fail. I'm currently unsure how to fix this. I've tried a few of the more advanced esi features, including ensuring cookies make it back to the origin server and things seem to work good. So, once the above bugs are figured out (particularly the include one), I think it will be in pretty good shape. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-1103) Traffic Server ESI plugin issues
[ https://issues.apache.org/jira/browse/TS-1103?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13208216#comment-13208216 ] Conan Wang commented on TS-1103: agree that gzip module of ESI is invalid for chrome. Traffic Server ESI plugin issues Key: TS-1103 URL: https://issues.apache.org/jira/browse/TS-1103 Project: Traffic Server Issue Type: Bug Components: Plugins Affects Versions: sometime Environment: Newest trunk. Reporter: Kevin Fox Attachments: esi.patch Patch to fix: * Makefile fix to add missing files. * Change return code checking to match whats trunk trafficserver. * Include missing header files. * Fix c++ namespace issues. * Work around strange name mangling/linking issue. * Force the assumption that the cached data is RAW_ESI, not PACKED_ESI. Things wouldn't work without it. * Comment out a block of code that looked to be incorrectly handling EOF. After this, simply loading the plugin and setting response header X-ESI in apache httpd seems to work. A few further bugs I have bumped into that aren't addressed in this patch: * It doesn't seem to parse gzip like it looks like it should. To work around, I had to disable it in apache httpd with RewriteRule . - [E=no-gzip:1] * If the client requests gzip, the ESI processor will gzip the result. It works in firefox but is invalid in chrome. Pulling a dump with curl and running it through gzip --list shows it has the correct uncompressed size and compressed size. using zcat shows the correct data but has the warning: invalid compressed data--length error. As far as I read the gzip spec though, the raw binary file looks valid to me. Not sure what this is. This can probably be simply disabled for now though. * esi:include is slightly broken. You get all the data back properly but sometimes the headers are sent prematurely with a Content-Length of 2**31-1. This causes clients to timeout and fail. I'm currently unsure how to fix this. I've tried a few of the more advanced esi features, including ensuring cookies make it back to the origin server and things seem to work good. So, once the above bugs are figured out (particularly the include one), I think it will be in pretty good shape. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira