Re: RAND_egd() blocking -- despite contract that states otherwise?

An alternative might be to install a handler for SIGALARM, and then
call alarm() to ensure that the blocked read returns with an EINTR?

-Kyle H

On Sun, Nov 9, 2008 at 2:06 PM, David Schwartz <davids@webmaster.com> wrote:[color=blue]
>[color=green][color=darkred]
>>> So what do you want to do if you run out of entropy?[/color][/color]
>[color=green]
>> Fail with an error condition stating that, rather than
>> the indeterminate hang in read() that was experienced.[/color]
>
> I believe you need to compile with EGD support then. This will get you the
> behavior you want. EGD provides no way to tell whether there's entropy or
> not, so if you fall back to it, and it has no entropy, you will be in
> trouble.
>
> There really is no way to fix this in OpenSSL. If you make it really not
> block, it will never succeed. It is meaningless to query a daemon without
> blocking -- at some point you must wait for the daemon to reply.
>
> DS
>
>
> ______________________________________________________________________
> OpenSSL Project [url]http://www.openssl.org[/url]
> User Support Mailing List [email]openssl-users@openssl.org[/email]
> Automated List Manager [email]majordomo@openssl.org[/email]
>[/color]
______________________________________________________________________
OpenSSL Project [url]http://www.openssl.org[/url]
User Support Mailing List [email]openssl-users@openssl.org[/email]
Automated List Manager [email]majordomo@openssl.org[/email]