rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-747307674
shouldnt the doc mention jaxrs too?
This is an automated message from the Apache Git Service.
To respond to the
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-739840070
Doc looks ok
(https://github.com/apache/cxf/pull/721/commits/72930d287d6e94e6ed742331bbf4f0d2ce645129#diff-d4ccfe1d5ea99f1cf8fc897e745c724ec4a2e2bdbff690a2f088989871f20e4bR21
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-739732809
Except ClassGeneratorClassLoader static variables which still look fishy and
unexpected - they hide another bug to be concrete like during redeployment of a
server with some
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-735285989
While there is a clear path for build time generation without quarkus and
(worse case) with a small custom main im fine with a smaller first step (we
should probably document it
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-735072209
@dufoli I'd really drop it since it has a lot of pitfall. You can have one
implementation which would use it as non static attribute and keep the default
one having a single map
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-733204047
@dufoli LC_ALL=en_US.UTF-8 ?
This is an automated message from the Apache Git Service.
To respond to the message,
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-732147376
@dufoli it is not needed, think a final "ultra clean" patch can be to force
visitMaxs(0, 0) call on visitEnd() one and drop all ref to visitMaxs but if it
works it is good for
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-728084199
@reta since "1." and "2." are not in this PR - future target - we are good
to merge the build time class generation code (it is the only feature in this
PR), right?
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727945948
@reta well, I see two options: do another pre-721 PR with the archi refacto
and then rebase 721 (but can be some work which will not bring much IMHO) or go
with 721 as it. a Next
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-72755
@dufoli if you can, at least the classloader one is important. A doc word on
how to capture the proxies at build time - even with a "main" and
exec-maven-plugin if you don't want
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727582333
2. yes but the classloader is not always the class classloader, it can be
another one (more or less for the same reason you use a classloader to store
proxies). Some servers
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727577291
@dufoli
1. cause the parameters are not used anyway and I started to investigate if
it was not the source of the stack error. Sure it can be in another patch (just
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727572118
@dufoli congrats ;)
Wonder if the static weak maps can't be dropped with something like:
diff --git
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727558031
@dufoli how do I reproduce the error? ran core asm tests and jaxb tests and
both run on my computer.
This is an
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727541323
@dufoli when you do a classwriter.toByteArray() (like
https://github.com/apache/cxf/pull/721/files#diff-0712fa792505d468314a10232664fa6b6d1c7cb1b58a8048b8310be1f677ae37R82)
you
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727536325
@dufoli if you can point out some code i can probably have a more precise
look. Classes being unique per classloader it can have been a bug which was not
preventing to work but
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727532121
You can debug normally the tests (through an IDE).
Now you can also dump the bytecode you generate in a /tmp/foo.class and open
it with jd-gui or javap to check what was
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727240024
@dufoli the bug refacto is to encapsulate asm, not this PR ;). IMHO adding
such a method was a blocker because it was only needed for your project and was
not achieving the goal
rmannibucau commented on pull request #721:
URL: https://github.com/apache/cxf/pull/721#issuecomment-727227166
@dufoli guess this class can stay static since it is a list of constants - I
suspect a clean refacto would be to encapsulate all these internals but this is
way too much for this
19 matches
Mail list logo
