On Wed, 4 Nov 2026 22:20:24 GMT, Anthony Scarpino <[email protected]> wrote:
> > Was there a reason for only adjusting this in the read cipher?
> > `SSLWriteCipher` appears to have the same issue:
> > ```
> >
> >
> > public boolean atKeyLimit() {
> >
> >
> > if (keyLimitCountdown >= 0) {
> >
> >
> > return false;
> >
> >
> > }
> >
> >
> >
> >
> >
> > // Turn off limit checking as KeyUpdate will be occurring
> >
> >
> > keyLimitEnabled = false;
> >
> >
> > return true;
> >
> >
> > }
> >
> >
> > ```
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > As with the read cipher before this patch, if `keyLimitCountdown` drops
> > below zero, `keyLimitEnabled` will be set to `false`. But this has no
> > effect on subsequent calls to this method which will again set the variable
> > to `false` and return `true` to say the limit has been reached, even though
> > the check should have been disabled by the first call.
>
> Testing never relieved the write-side to be a problem because it depended on
> the read-side being open. If you are able to create a test and show there is
> a problem, please file a follow-on bug and make a PR. I can review it. Thanks.
I haven't seen a problem, just curious about the asymmetry between the two.
[Francisco](https://github.com/openjdk/jdk21u-dev/pull/3106#issuecomment-6614408897)
gave a explanation of why we probably won't see the same on the write side,
which I'm happy with, on the 21u PR. Thanks.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/30288#issuecomment-6811975453
