hi!
i started to receive errors from poll due to EAGAIN and it broke the CGI handling in our web server. the issue is not urgent, but i thought i'll report it.
if i add the following to PROCESS-POLL-REVENTS under the case (ISYS:POLLERR):
(error 'poll-error :fd fd :identifier (princ-to-string (iolib.syscalls:errno))))
then i see 11, which is EAGAIN.
shouldn't that be handled by iolib behind my back? or am i doing somehting wrong?
the sympthoms are that the request is not even parsed properly (iiuc), but when i use View Source in chrome then it seems to work. it's maybe something about the request not arriving in one packet or somesuch? but my network foo is limited...
the code in question is in hu.dwim.web-server.
On Fri, 2014-06-20 at 01:18 +0200, Attila Lendvai wrote:
hi!
i started to receive errors from poll due to EAGAIN and it broke the CGI handling in our web server. the issue is not urgent, but i thought i'll report it.
if i add the following to PROCESS-POLL-REVENTS under the case (ISYS:POLLERR):
(error 'poll-error :fd fd :identifier (princ-to-string (iolib.syscalls:errno))))
then i see 11, which is EAGAIN.
shouldn't that be handled by iolib behind my back? or am i doing somehting wrong?
the sympthoms are that the request is not even parsed properly (iiuc), but when i use View Source in chrome then it seems to work. it's maybe something about the request not arriving in one packet or somesuch? but my network foo is limited...
the code in question is in hu.dwim.web-server.
Hi Attila, if you can grab a stacktrace, it would be very useful. I thought I handled all places where EAGAIN could be triggered, but obviously I haven't.
i started to receive errors from poll due to EAGAIN and it broke the CGI handling in our web server. the issue is not urgent, but i thought i'll report it.
if i add the following to PROCESS-POLL-REVENTS under the case (ISYS:POLLERR):
(error 'poll-error :fd fd :identifier (princ-to-string (iolib.syscalls:errno))))
then i see 11, which is EAGAIN.
shouldn't that be handled by iolib behind my back? or am i doing somehting wrong?
the sympthoms are that the request is not even parsed properly (iiuc), but when i use View Source in chrome then it seems to work. it's maybe something about the request not arriving in one packet or somesuch? but my network foo is limited...
the code in question is in hu.dwim.web-server.
Hi Attila, if you can grab a stacktrace, it would be very useful. I thought I handled all places where EAGAIN could be triggered, but obviously I haven't.
FTR, i cannot reproduce this anymore.
i've updated iolib to the latest on dwim.hu. i didn't update libfixposix because this repo is gone:
http://download.opensuse.org/repositories/home:/sionescu/Debian_Squeeze/./Pa...
then i tried the darcsweb.cgi with chrome and it works. so, no backtrace... :)
http://dwim.hu/darcsweb/darcsweb.cgi
-
Attila Lendvai -
Attila Lendvai -
Stelian Ionescu