Re: curl exit code change after 335dc0e3c5968827 intentional?

Fabian Keil via curl-library <[email protected]> wrote on 2026-05-13 at 14:53:24:

> Fabian Keil via curl-library <[email protected]> wrote on 2026-04-15 at 15:31:35:
> 
> > Daniel Stenberg <[email protected]> wrote on 2026-04-13 at 19:28:26:
> > 
> > > On Mon, 13 Apr 2026, Fabian Keil via curl-library wrote:
> > > 
> > > > test 0004...[Request from client address from denied network range]
> > > >
> > > > curl returned 56, when expecting 52
> > > 
> > > Clearly your test case now fails differently, so it gets another error code 
> > > because 52 is still there and returned for the same reason. We even have test 
> > > cases for it.
> > > 
> > > It would be helpful if you could make a test case with the curl test suite 
> > > that reproduces the problem you experience.
> > 
> > The patch set at [0] adds a test server that drops connections
> > without returning data and adds two tests that use it.
> > 
> > curl versions before 335dc0e3c report an "Empty reply from server",
> > afterwards its a "Failure in receiving network data".
[...]
> Should I submit a PR to get a review or is the patch considered not worth it?

A couple of days after I modified two Privoxy tests to expect error code 56
from curl instead of 52 [1] it was reported that those modified tests now
actually fail on Debian GNU/Linux [2] where curl continues to return 52.

I was able to reproduce this and as it turns out the two tests added
by the patch set in [0] fail on GNU/Linux as well, but work on ElectroBSD
(and therefore probably FreeBSD) and OpenBSD.

Fabian

[0]: <https://www.fabiankeil.de/sourcecode/curl-connection-dropper-tests.diff>
[1]: <https://www.privoxy.org/gitweb/?p=privoxy.git;a=commitdiff;h=cd39f62455ce4808ae7b89bb82f08765c42ac8dd>
[2]: <https://lists.privoxy.org/pipermail/privoxy-devel/2026-May/000849.html>
-- 
Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library
Etiquette:   https://curl.se/mail/etiquette.html