quickfix-developers

[Quickfix-developers] QuickFIX, java and threads - added blockingStart() to SocketAcceptor

[Alex H. <al...@an...>]

Hi,

On my platform (redhat 8,  gcc 3.2, sun jdk 1.4.1_02) I'm still getting
thread crashes if threads are spawned outside java (perhaps a thread
local storage problem in the JVM?).

So as to not affect the behaviour of the existing start() method,
instead of modifying start(), I have added a blockingStart() method to
the SocketAcceptor.

This fixes my crash bug, and may also be useful to people that want to
handle FIX messages in the main thread (e.g. if FIX is all the
application does....)

ChangeLog follows.

Cheers,
Alex.

2003-05-06 Alex Hornby <al...@an...>

	* src/java/org_quickfix_SocketAcceptor.cpp: add blockingStart

	* src/java/org_quickfix_SocketAcceptor.h: add blockingStart

	* src/java/src/org/quickfix/SocketAcceptor.java: add blockingStart

Re: [Quickfix-developers] QuickFIX, java and threads - added blockingStart() to SocketAcceptor

[Joerg T. <Joe...@ma...>]

Alex Hornby wrote:
> Hi,
> 
> On my platform (redhat 8,  gcc 3.2, sun jdk 1.4.1_02) I'm still getting
> thread crashes if threads are spawned outside java (perhaps a thread
> local storage problem in the JVM?).

Interesting. We have a similiar setup (debian 2.2, gcc 3.0, sun jdk 1.4.1_01),
and no problems so far. Currently, we use the SocketAcceptor. How does these
crashes show up?

Jörg

-- 
Joerg Thoennes
                                http://macd.com
Tel.: +49 (0)241 44597-24      Macdonald Associates GmbH
Fax : +49 (0)241 44597-10      Lothringer Str. 52, D-52070 Aachen

Re: [Quickfix-developers] QuickFIX, java and threads - added blockingStart() to SocketAcceptor

[Alex H. <al...@an...>]

On Tue, 2003-05-06 at 14:29, Joerg Thoennes wrote:
> Alex Hornby wrote:
> > Hi,
> >=20
> > On my platform (redhat 8,  gcc 3.2, sun jdk 1.4.1_02) I'm still getting
> > thread crashes if threads are spawned outside java (perhaps a thread
> > local storage problem in the JVM?).
>=20
> Interesting. We have a similiar setup (debian 2.2, gcc 3.0, sun jdk 1.4.1=
_01),
> and no problems so far. Currently, we use the SocketAcceptor. How does th=
ese
> crashes show up?
>=20
> J=F6rg

fromAdmin() gets a null pointer back as the java environment, which then
causes a crash as soon as it is used...

Alex.