Re: HTTP2 parsing and error handling

2015-06-10 Thread Mark Thomas
On 09/06/2015 11:31, Rémy Maucherat wrote:
 2015-06-09 11:53 GMT+02:00 Konstantin Kolinko knst.koli...@gmail.com:
 
 2015-06-08 15:37 GMT+03:00 Mark Thomas ma...@apache.org:
 All,

 I'm not particularly happy with the current error handling (based around
 Http2Exception) and as I work my way through the spec in detail I am
 coming across cases that are becoming increasingly difficult to handle
 cleanly.

 I think I have reached the point where a completely different approach
 is needed. I am thinking along the lines of a ParseState object that
 gets passed around and updated if things go wrong. I'm going to try this
 approach over the next few days. If it works there is likely to be a
 largish commit to switch to this new approach but I'll try and keep that
 commit limited to changing the error handling.

 +1

 One bonus from a flag vs. an exception is that it can be sticky.
 I mean a situation like one fixed by this commit:
 http://svn.apache.org/r1600984

 
 Yes, and then it doesn't get properly recycled :) That aside, flags are
 often good.

I managed to clean things up and make progress without such a large
refactoring. I haven't finished implementing all the checks yet so I may
still come back to this approach if the code starts to get messy again.

Mark


-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



Re: HTTP2 parsing and error handling

2015-06-09 Thread Konstantin Kolinko
2015-06-08 15:37 GMT+03:00 Mark Thomas ma...@apache.org:
 All,

 I'm not particularly happy with the current error handling (based around
 Http2Exception) and as I work my way through the spec in detail I am
 coming across cases that are becoming increasingly difficult to handle
 cleanly.

 I think I have reached the point where a completely different approach
 is needed. I am thinking along the lines of a ParseState object that
 gets passed around and updated if things go wrong. I'm going to try this
 approach over the next few days. If it works there is likely to be a
 largish commit to switch to this new approach but I'll try and keep that
 commit limited to changing the error handling.

+1

One bonus from a flag vs. an exception is that it can be sticky.
I mean a situation like one fixed by this commit:
http://svn.apache.org/r1600984

Best regards,
Konstantin Kolinko

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org



Re: HTTP2 parsing and error handling

2015-06-09 Thread Rémy Maucherat
2015-06-09 11:53 GMT+02:00 Konstantin Kolinko knst.koli...@gmail.com:

 2015-06-08 15:37 GMT+03:00 Mark Thomas ma...@apache.org:
  All,
 
  I'm not particularly happy with the current error handling (based around
  Http2Exception) and as I work my way through the spec in detail I am
  coming across cases that are becoming increasingly difficult to handle
  cleanly.
 
  I think I have reached the point where a completely different approach
  is needed. I am thinking along the lines of a ParseState object that
  gets passed around and updated if things go wrong. I'm going to try this
  approach over the next few days. If it works there is likely to be a
  largish commit to switch to this new approach but I'll try and keep that
  commit limited to changing the error handling.

 +1

 One bonus from a flag vs. an exception is that it can be sticky.
 I mean a situation like one fixed by this commit:
 http://svn.apache.org/r1600984


Yes, and then it doesn't get properly recycled :) That aside, flags are
often good.

Rémy


HTTP2 parsing and error handling

2015-06-08 Thread Mark Thomas
All,

I'm not particularly happy with the current error handling (based around
Http2Exception) and as I work my way through the spec in detail I am
coming across cases that are becoming increasingly difficult to handle
cleanly.

I think I have reached the point where a completely different approach
is needed. I am thinking along the lines of a ParseState object that
gets passed around and updated if things go wrong. I'm going to try this
approach over the next few days. If it works there is likely to be a
largish commit to switch to this new approach but I'll try and keep that
commit limited to changing the error handling.

Mark

-
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org