Fix potential Null pointer dereference in regress_ws.c

[icy17]

Fix #1604

[icy17]

I'm not sure if I should close fd before tt_assert...

[azat]

I'm not sure if I should close fd before tt_assert...

Yes, it should be done under end label, and you need to initialize fd with -1

[azat]

.

[azat]

if (fd != EVUTIL_INVALID_SOCKET) (and use it for initilzation), and also let's add a check for http

[icy17]

I noticed that the http is not used here? I'm not quite sure where I'm supposed to add the check.
Additionally, I should initialize fd to EVUTIL_INVALID_SOCKET and check if it is EVUTIL_INVALID_SOCKET before closing, is that right?

[azat]

I noticed that the http is not used here?

It is used it creates http server, see usage of port.

I'm not quite sure where I'm supposed to add the check.

The http initialization may fail, so - if (!http) goto err

Additionally, I should initialize fd to EVUTIL_INVALID_SOCKET and check if it is EVUTIL_INVALID_SOCKET before closing, is that right?

Yes

[azat]

/home/runner/work/libevent/libevent/test/regress_ws.c:351:2: error: ISO C90 forbids mixed declarations and code [-Werror=declaration-after-statement]
  351 |  struct evbuffer *out;
      |  ^~~~~~

[Coeur]

There seems to be a lot of failures with:

[warn] evbuffer_file_segment_materialize: mmap(5, 0, 0) failed: Bad file descriptor

Could those failures be related to those changes (more specifically, the if (!http) goto end;)?

[azat]

Please resubmit