[tin-dev] [PATCH] [WIP] NNTPS support

[Urs Janßen]

On Sat, Dec 05, 2020 at 04:26:02PM +0000, Pierre-Alexandre Meyer wrote:

First of all,
thanks for showing interest in that point :)

> This is still very much a work in progress (I mostly only tested batch
> mode), but I wanted

interactive connection looks ok (snipped a bit of the data as it's not my
server):

<<< [18:47:11.883317] (TLS1.2)-(ECDHE-RSA-SECP256R1)-(AES-256-GCM)
<<< [18:47:16.962407] 200 news[...]
>>> [18:47:16.962640] CAPABILITIES
<<< [18:47:17.025975] 101 Capability list:
[...]
<<< [18:47:17.081551] SASL SCRAM-SHA-1 GS2-IAKERB GS2-KRB5 GSSAPI GSS-SPNEGO DIGEST-MD5 OTP NTLM CRAM-MD5 LOGIN PLAIN
>>> [18:47:17.082016] LIST MOTD
<<< [18:47:17.145823] 480 Authentication required for command
>>> [18:47:17.147620] AUTHINFO SASL PLAIN AFVycwAgdGVzdCA=
<<< [18:47:17.212225] 481 generic failure

(AFVycwAgdGVzdCA= is base64("\0username\0password") incl. the
(leading/tailing) spaces the password has, might be an issue on
the servers side.)

>>> [18:47:19.213218] AUTHINFO USER [...]
<<< [18:47:19.274362] 381 Enter password
>>> [18:47:19.274666] AUTHINFO PASS [...] 
<<< [18:47:19.346595] 281 Authentication succeeded
>>> [18:47:21.347515] CAPABILITIES
[...]

but once you enter a group it get stuck in

>>> [19:05:30.440199] LISTGROUP [...].test
<<< [19:05:30.518678] 211 total low high [...].test
>>> [19:05:30.542676] OVER low-high
<<< [19:05:30.623157] 224 Overview information for low-high follows
[... overview data ...]
.
^ that lone extra dot after the last overview record is unecpcted

#5  0x000000000044af07 in get_server (string=string at entry=0x7f9f2c92b010 "", 
    size=size at entry=1634048) at ./nntplib.c:1036
#6  0x0000000000466a4b in tin_read (buffer=0x7f9f2c92b010 "", len=1634048, 
    fp=fp at entry=0x270f, header=header at entry=0) at ./read.c:219
#7  0x0000000000466757 in tin_fgets (fp=0x270f, header=0) at ./read.c:347
#8  0x000000000040a572 in read_overview (group=0x5, 
    group at entry=0x7f9f348b8990, min=35900032, min at entry=1, max=5, 
    max at entry=14170, top=0x7f9f3302747d <__libc_recv+29>, 
    top at entry=0x7ffd9bd74b10, local=local at entry=0, rebuild_cache=0x0, 
    rebuild_cache at entry=0x7ffd9bd74b0c) at ./art.c:1946
#9  0x000000000040d708 in index_group (group=group at entry=0x7f9f348b8990)
    at ./art.c:482
#10 0x000000000042b5bc in group_page (group=0x7f9f348b8990) at ./group.c:171
#11 0x00000000004755b9 in read_groups () at ./select.c:1216
#12 0x00000000004756d1 in select_read_group () at ./select.c:1422
#13 0x0000000000476df5 in selection_page (
    start_groupnum=start_groupnum at entry=0, 
    num_cmd_line_groups=num_cmd_line_groups at entry=0) at ./select.c:231
#14 0x0000000000405772 in main (argc=<optimized out>, argv=<optimized out>)
    at ./main.c:436

> to reach out early for feedback. Specifically, how do you feel about
> two codepaths for
> the I/O code (e.g. s_gets vs gnutls_record_recv)?

I have no objection to that

> Notes:
> 
> * To enable the feature, use the new --enable-gnutls configure flag at
> build time.
> * Use the new -k runtime flag to force TLS.

which should probably also change the default port to 563 as long as no
-p flag is present (and/or optionally take a port).

> * I didn't submit the configure script changes, as I'm running a
> different toolchain (GNU Autoconf 2.69).

JFTR: I use Thomas E. Dickey's version:
<https://invisible-island.net/autoconf/>
<https://invisible-island.net/datafiles/release/autoconf.tar.gz>