tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rajiv Mordani <m...@chinet.com>
Subject Re: nightly build scripts (was: IllegalAccessError and Tags)
Date Thu, 03 Feb 2000 21:27:03 GMT
The bug that is there in the compiler is that all the calls to methods in
the byte code generated by a 1.1 are fully qualified.  So what happens is that 
for a call to a public method in ParentNode the byte code generated has the 
call to ParentNode.publicMethod() and not just publicMethod () which causes 
the IllegalAccessError. This was changed for 1.2 and later. However it was a 
regression in the 1.3 ea release (has been fixed for the next release).

- Rajiv

Danno Ferrin wrote:
> 
> 
> rubys@us.ibm.com wrote:
> > 
> > Danno wrote:
> > > Is this being caused by old or dirty builds visible in the class path?
> >
> > Absolutely nothing was in the class path.  Everything was built from a
> > clean cvs checkout.  (not even an update, a clean checkout)
> 
> That is good to hear.  And since it is 1.1.8 nothing can hide in the 
> 
> > According to a prior post by Rajiv on this topic, apparently the Java
> > compilers with 1.1 and 1.3 are not compatible with 1.2, and the consensus
> > is that the 1.2 behavior is correct and this is being pursued as a bug in
> > 1.3 (but curiously not being fixed in 1.1).
> 
> I would be interested to know what the particular compiler bug is.  But
> I think sun's policy that a known bug with a known fix will not be
> rolled into a 1.2.3 or 1.1.9 without a $100,000 fee is appalling, this
> is the explaination I got when I reported the previously mentioned "try
> nothing" verification error bug.
> 
> Perhaps we should consider moving to a different compiler like jikes or
> the 1.3 compiler for the distros.  The 1.3 part would nix compilation on
> linux but with an "at" job and an FTP pull or a samba mount from an nt
> box it could be automated.  
> 
> > My temptation is to produce separate builds for each major JDK level in the
> > 3.1_M2 timeframe.
> 
> This may be best.  I haven't run a full matrix of source compilers and
> jvm/oses but it might prove interesting if I ever get the time.
> 
> > - Sam Ruby
> > 
> > P.S.  At this point, I am having deep difficulty in repressing my Italian
> > heritage, and the impulse to express my outrage in a very forceful and
> > direct manner.  Perhaps Stefano can help me out here when he gets back on
> > Friday?  <grin>
> 
> Too many feature requests/demands and not enough developer bandwidth can
> do that to anyone.
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: tomcat-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: tomcat-dev-help@jakarta.apache.org
> 


-- 
  UNIX _is_ user friendly, 
  he's just very picky about who his friends are.

Mime
View raw message