httpd-bugs mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject [Bug 57045] dumpio output strings are missing
Date Tue, 06 Oct 2015 00:37:20 GMT

--- Comment #8 from Kurtis Rader <> ---
> This patch looks not thread-safe...

It's definitely not thread-safe. I should have mentioned that this version only
works reliably with MPM pre-fork; which is what I'm currently using since my
site runs WordPress. I threw this together over a couple of evenings and wanted
to get it out there for feedback. Assuming no one has any other problems with
the code I'll add the necessary support for thread local data to make it

> How about using ap_log_[r]data() introduced in 2.4.11?

That looks like it will emit data in a non-backward compatible format. It would
also mean the code can't be used with 2.2.x. For this change I'd prefer to be
conservative and keep the logged format unchanged to the extent possible and
remain compatible with 2.2.x. Obviously I'm changing how 0xff is represented in
the error log but it's a safe bet no one will care because they've obviously
never tried to use the binary data that is logged.

P.S., I looked at the existing uses of ap_log_rdata() in the 2.4.16 tree and
they all appear to be incorrect. The ap_log_rdata function declaration doesn't
match any of the uses. Consider the sole use in modules/proxy/mod_proxy_fcgi.c:

    ap_log_rdata(APLOG_MARK, APLOG_TRACE8, r, "FastCGI header",
                 farray, AP_FCGI_HEADER_LEN, 0);

The only reason this compiles is because "r" is defined at function scope. If
you rename "r" to "req" throughout the function it fails to compile with this

mod_proxy_fcgi.c:577:13: error: use of undeclared identifier 'r'
            ap_log_rdata(APLOG_MARK, APLOG_TRACE8, req, "FastCGI header",
/Users/krader/brew-httpd-2.4.16/include/http_log.h:636:27: note: expanded from
macro 'ap_log_rdata'

P.P.S., That magic zero as the last parameter of the call rather than

You are receiving this mail because:
You are the assignee for the bug.

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message