On Saturday 24 June 2006 21:14 Mark Hindess wrote: > On 24 June 2006 at 20:10, Gregory Shimansky <[EMAIL PROTECTED]> wrote: > > On Saturday 24 June 2006 19:16 Geir Magnusson Jr wrote: > > > I seem to have found a doozy :) > > > > > > In hysock.c / getNextNetlinkMsg() there's an assumption that anything > > > the kernel will ever send in a netlink request is less than 1k, so when > > > processing the receive buffer, if there still is data in the buffer and > > > it's not a complete message, it's considered an error. > > > > > > This is clearly wrong, and provable by simply doubling the buffer size. > > > This is suboptimal, of course, because there is no guarantee that 2k > > > will be enough one all systems either. > > > > > > I'm going to stare at this and attempt to rewrite it this weekend if I > > > find some time. > > > > > > Man, this would be easier w/ a debugger :) > > > > I have a question on this topic. Is there a reason why libhy*.so and > > other classlib natives libraries are compiled without debug information? > > Addind -g (-Zi on windows) to the options could really help sometimes. > > No. At least there isn't a good reason. We should fix this. I've > added it manually a couple of times which is always a sign that > something is needs fixing.
To make it happen sooner rather than later I decided to create a JIRA HARMONY-681 with the changes which I've made locally some time ago already. It enables debug symbols info for classlib (not touching -O1 optimization flag so debuging may still be inaccurate at times) and full debug mode for drlvm by default (Geir asked about how to do it some time ago). If I've missed something in the middle of classlib native code revamp please update the patch or let me know. -- Gregory Shimansky, Intel Middleware Products Division --------------------------------------------------------------------- Terms of use : http://incubator.apache.org/harmony/mailing.html To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
