Re: gnutls problems again

[Ted Zlatanov <tzz@lifelogs.com>]

On Tue, 15 Mar 2011 22:19:09 +0100 Lars Magne Ingebrigtsen <larsi@gnus.org> wrote: 

LMI> The Gmail IMAP servers seem to be having problems again, but it triggers
LMI> an interesting gnutls-related bug.

LMI> If I say:

LMI> (gnus-open-server '(nnimap "imap.gmail.com"))

LMI> then Emacs starts using 100% CPU.  Emacs doesn't hang, but as long as
LMI> that connection is up (i.e., until I kill the process buffer), then
LMI> strace says the following:

LMI> poll([{fd=4, events=POLLIN}, {fd=3, events=POLLIN}], 2, 0) = 0 (Timeout)
LMI> select(10, [3 4 6 7 8], [], NULL, {0, 0}) = 1 (in [8], left {0, 0})
LMI> getrusage(RUSAGE_SELF, {ru_utime={44, 494780}, ru_stime={8, 100506}, ...}) = 0
LMI> times({tms_utime=4449, tms_stime=810, tms_cutime=0, tms_cstime=0}) = 1731291756
LMI> getrusage(RUSAGE_SELF, {ru_utime={44, 494780}, ru_stime={8, 100506}, ...}) = 0
LMI> times({tms_utime=4449, tms_stime=810, tms_cutime=0, tms_cstime=0}) = 1731291756
LMI> read(3, 0x11fe4f4, 4096)                = -1 EAGAIN (Resource temporarily unavailable)
LMI> poll([{fd=4, events=POLLIN}, {fd=3, events=POLLIN}], 2, 0) = 0 (Timeout)
LMI> select(10, [3 4 6 7 8], [], NULL, {0, 0}) = 1 (in [8], left {0, 0})
LMI> getrusage(RUSAGE_SELF, {ru_utime={44, 494780}, ru_stime={8, 100506}, ...}) = 0
LMI> times({tms_utime=4449, tms_stime=810, tms_cutime=0, tms_cstime=0}) = 1731291756

LMI> And a *lot* of that.

Claudio Bley's patch (posted on emacs-devel recently) does some EAGAIN
reworking which may fix this.  His code is for W32 but it has general
fixes too.  He's supposed to look at it again this weekend and maybe
bring in my crappy half-baked patch for callback support, which lets you
set the verify flags.  So maybe you want to work with him on debugging
this hang.

Ted