On May 31, 2014, at 6:08 PM, Brendan MacDonell <[email protected]> wrote:

> On Sat, May 31, 2014 at 8:02 PM, Brent Cook <[email protected]> wrote:
> I totally agree in the general case. BIO is a big pain, and it does seem 
> crazy to use it for stdio.
> 
> However, in this specific case, this file already calls BIO_printf, BIO_puts 
> and BIO_write to stdout and stderr, in an unchecked manner, several hundred 
> times. I’m not sure if checking write() actually fixes the problem, as there 
> may be no guarantees of IO ordering when mixing BIO_write, BIO_puts and 
> regular write() calls to the same output stream.
> 
> Might it be OK to consider unifying this single exception to the (bad) rule, 
> and then removing all BIO to stdin/stdout all at once in a follow-up patch?
> 
> For what it's worth, bio_s_out is not stdout if -quiet is passed (it's a 
> BIO_s_null instead.) If you change this write to BIO_write, then it will be 
> suppressed as well.
> 


Good point. The help for this command says:

   -quiet        - No server output

It seems like supressing here would be the correct behavior as well.

Maybe a larger question is, should the openssl command really embed a 
static-file-serving HTTP 1.0 web server? The ‘-hack’ and ‘-bugs’ options seem 
like a little ripe, among many other things.

 - Brent

Reply via email to