On 19.01.2016, at 02:36, Yasuo Ohgaki <yohgaki@ohgaki.net> wrote:
>
> Hi David,
>
> On Tue, Jan 19, 2016 at 2:31 AM, David Zuelke <dz@heroku.com> wrote:
>>
>> I have customers running into this too. I have not yet had time to pick up that patch - AFAICT, it's incomplete (see comments).
>>
>> We are, in parallel, working on getting a fix into libpq, but that's of course just a band-aid, since it can still happen with any other library that also uses openssl functions.
>
> That's what I thought.
>
>>
>> The principle approach of that PR ("backing up" the errors from openssl) is good, though.
>
> This issue is nasty. I hope you come up with feasible solution soon.
The libpq fix is in the works and not that complex. It's more a question of when it would get released (Postgres doesn't do regular release intervals).
The PHP side of it is more work, and genuinely not pretty, but it's not like there's an alternative approach to fix it, so we'll have to live with it.
What I am currently struggling with is a way to reproduce this for the tests... once libpq is fixed, the problem can't be triggered anymore.
> BTW, should I change the bug report category to openssl?
Yeah it's an ext-openssl "problem" (one could also argue the fault is with OpenSSL, but meh).
David
