matt massie wrote:
i've just checked in changes to ./gmetad/process_xml.c . the
process_xml() function was freeing memory for xml_data->cluster and
xml_data->host and was overwriting the struct with a memset() zero.
also, an XML error would prevent the XML parser from being freed.
# cvs update ./gmetad/process_xml.c
is recommended.
This seems to have helped metric source stability on this end somewhat, but
now gmetad occasionally segfaults (still getting that "expecting one data
source, 0 returned" error every now and then). I've checked the data and
it looks normal. No funky characters.
I am going to crack open Expat to see if I can get a more verbose
description of exactly what the XML parser is having trouble with, in the
hopes of discovering a smoking gun.
