On Wed, Jan 27, 2021 at 10:16 PM Alexandr Miloslavskiy
<alexandr.miloslavs...@syntevo.com> wrote:
>
> Thanks for reviewing! You're quite the savior of this branch :)
Heh, you're welcome :). And thanks for bringing it here in the first place.
> On 27.01.2021 11:02, Johan Corveleyn wrote:
> > was wondering what you meant by "This error is easily seen
> > when running JavaHL tests"
>
> If you checkout r1882518, tests die with 'Bad global or local ref passed
> to JNI' before this error can be seen. However, if you also cherry-pick
> r1882522 on top to fix this problem, that's where you'll see 'JNI call
> made with exception pending' that I fixed in r1882520.
Ah, okay. I'll give that another try.
> Sorry for confusion. I edited/reordered patches a few times before
> committing, and didn't notice that the commit message is not too good
> with this ordering.
>
> Which brings me to next question. In other replies, you suggested that I
> make some edits in code and also amend commit messages. How do I do that
> in SVN? I understand that unlike git, a submitted branch is already set
> in stone? Please advise. This branch is my sole SVN experience in entire
> life, so I'm quite a newbie. I did google and ask another guy, though,
> to no avail.
Well, committed content is indeed "set in stone", so to speak (i.e.
"part of history"). You could of course start a new branch and redo
everything, but I wouldn't recommend it (that would force me / others
to review it all again). The usual practice, if there are remarks
about the content, is to simply add additional commits fixing the
problems / adding more code / ... So for changes to the contents of
the file, I suggest you perform further commits on the branch.
OTOH, _commit messages_ are editable in SVN. They are part of the so
called "revision properties" (contrary to "versioned properties"). So,
for remarks about the commit messages, the usual practice is to simply
edit them. For example with this command:
svn propedit --revprop -r$REV svn:log $URL
(the $URL can be any url pointing to the svn repository, since
revision numbers are global for the repo)
This will open the current log message in your $SVN_EDITOR (fallback
to $EDITOR).
> > I do get 8 warnings with this sort of message (always similar
> > stacktrace, always coming from UtilTests.testFileMerge, line 120):
> > [[[
> > WARNING: JNI local refs: 33, exceeds capacity: 32
> > at java.net.NetworkInterface.getAll(java.base@11.0.6/Native Method)
> > at
> java.net.NetworkInterface.getNetworkInterfaces(java.base@11.0.6/NetworkInterface.java:359)
>
> To my knowledge, this is an oversight in
> 'java.net.NetworkInterface.getAll' itself.
>
> JNI documentation says [1]:
> [[[
> You can call the JNI EnsureLocalCapacity() method to tell the JVM that
> you'll be using more than 16 local references. This allows the JVM to
> optimize the handling of local references for that native. Failure to
> inform the JVM can lead to a FatalError if the required local references
> cannot be created, or poor performance that's due to a mismatch between
> the local-reference management employed by the JVM and the number of
> local references used.
> ]]]
>
> I'm observing the same warnings in other Java programs which are
> unrelated to SVN. I think that this can be ignored for the purposes of
> this JavaHL related discussion.
>
> [1] https://developer.ibm.com/languages/java/articles/j-jni/
Yes, let's ignore that here ... it's unrelated.
--
Johan
