I just realized that [netsend] and [netreceive] are unreliable, and this may be the source of most of my crashes now.
When bursts of large numbers of messages are sent, there are random message losses. I may send 1000 messages in a burst and lose messages 613 through 918, with nobody issuing any error message or warning. Btw I am using TCP netsend/receive, not UDP. When a bunch of messages is lost, usually a corrupt message is received (i.e. netreceive outputs a garbage message that has never been sent: probably resulting from a "piece" of message). Among other things, I am using received messages to dynamically create objects in a subpatch, so god knows what happens when a garbage message is appended to a "obj x y" message sent to a canvas: that may be one source of crashes. Another one may be some symbol being sent to a numeric [select] or viceversa, or some message of an unexpected or illegal type entering an inlet it shouldn't enter. I am carefully checking and routing MY messages, but always assuming that only the kind of messages that I know are being sent can be received. TCP communication is all about guaranteing such an assumption is true. 2008/2/12, Miller Puckette <[EMAIL PROTECTED]>: > It's a dangerous thing to offer, but if you can make a patch crash > using 'vanilla' pd on any of 32-bit linux, windows, mac, or 64-bit > linux, send me a tarball of the patch and its dependencies and I'll > try to see what happened. Fixing crash bugs is always my highest priority. > > cheers > Miller > > > On Tue, Feb 12, 2008 at 03:46:11PM +0100, Steffen Juul wrote: > > > > On 12/02/2008, at 13.58, matteo sisti sette wrote: > > > > > Are there any changelogs (is this the correct word) listing added > > > features and especially bugfixes of each PD release? > > > > There are release notes in doc/1.manual/x5.htm. The bug/patch tracker > > holds some info, but are not complete either - not everything is > > listed there basically since Miller need not add to it but only pull > > from it, and not everything is guaranteed to be indexed/categorized > > right which means it's lost wrt. searching (like a misplaced book in > > the library). > > > > To get a complete list I'm afraid we are left we diff'ing the source. > > > > > > > > _______________________________________________ > > PD-list@iem.at mailing list > > UNSUBSCRIBE and account-management -> > > http://lists.puredata.info/listinfo/pd-list > -- Matteo Sisti Sette [EMAIL PROTECTED] http://www.matteosistisette.com _______________________________________________ PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list