10:33 a.m.
2010/4/2 Chris Lalancette <clalance(a)redhat.com>:
On 04/02/2010 09:25 AM, Matthias Bolte wrote:
> 2010/4/2 Chris Lalancette <clalance(a)redhat.com>:
>> Signed-off-by: Chris Lalancette <clalance(a)redhat.com>
>> ---
>> src/conf/domain_conf.c | 402 ++++++++++++++++++++++++++++++++++++++++++++--
>> src/conf/domain_conf.h | 53 ++++++
>> src/libvirt_private.syms | 10 ++
>> 3 files changed, 455 insertions(+), 10 deletions(-)
>>
>> diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
>> index e260dce..1971b9a 100644
>> --- a/src/conf/domain_conf.c
>> +++ b/src/conf/domain_conf.c
>
>> +virDomainSnapshotDefPtr virDomainSnapshotDefParseString(const char *xmlStr,
>> + int newSnapshot)
>> +{
>> + xmlXPathContextPtr ctxt = NULL;
>> + xmlDocPtr xml = NULL;
>> + xmlNodePtr root;
>> + virDomainSnapshotDefPtr def = NULL;
>> + virDomainSnapshotDefPtr ret = NULL;
>> + char *creation = NULL, *state = NULL;
>> + struct timeval tv;
>> + struct tm time_info;
>> + char timestr[100];
>> +
>> + xml = virXMLParse(NULL, xmlStr, "domainsnapshot.xml");
>> + if (!xml) {
>> + virDomainReportError(VIR_ERR_XML_ERROR,
>> + "%s",_("failed to parse snapshot xml
document"));
>> + return NULL;
>> + }
>> +
>> + if ((root = xmlDocGetRootElement(xml)) == NULL) {
>> + virDomainReportError(VIR_ERR_INTERNAL_ERROR,
>> + "%s", _("missing root
element"));
>> + goto cleanup;
>> + }
>> +
>> + if (!xmlStrEqual(root->name, BAD_CAST "domainsnapshot")) {
>> + virDomainReportError(VIR_ERR_INTERNAL_ERROR,
>> + "%s", _("incorrect root
element"));
>> + goto cleanup;
>> + }
>> +
>> + ctxt = xmlXPathNewContext(xml);
>> + if (ctxt == NULL) {
>> + virReportOOMError();
>> + goto cleanup;
>> + }
>> +
>> + if (VIR_ALLOC(def) < 0) {
>> + virReportOOMError();
>> + goto cleanup;
>> + }
>> +
>> + ctxt->node = root;
>> +
>> + def->name = virXPathString("string(./name)", ctxt);
>> + if (def->name == NULL) {
>> + /* make up a name */
>> + gettimeofday(&tv, NULL);
>> + localtime_r(&tv.tv_sec, &time_info);
>> + strftime(timestr, sizeof(timestr), "%F_%T", &time_info);
>> + def->name = strdup(timestr);
>> + }
>> + if (def->name == NULL) {
>> + virReportOOMError();
>> + goto cleanup;
>> + }
>> +
>> + def->description = virXPathString("string(./description)",
ctxt);
>> +
>> + if (!newSnapshot) {
>> + creation = virXPathString("string(./creationTime)", ctxt);
>
> I think it should be creationtime or creation_time, but not creationTime.
Taking the domain XML as an example, all 3 styles are used (e.g. currentMemory,
on_poweroff, and seclabel). I don't really care too much, so I'll do whatever
is more comfortable.
I missed currentMemory and since it's already a mixture of all styles
lets just keep creationTime.
>> +char *virDomainSnapshotDefFormat(char *domain_uuid,
>> + virDomainSnapshotDefPtr def)
>> +{
>> + virBuffer buf = VIR_BUFFER_INITIALIZER;
>> + char timestr[100];
>> + struct tm time_info;
>> +
>> + virBufferAddLit(&buf, "<domainsnapshot>\n");
>> + virBufferVSprintf(&buf, " <name>%s</name>\n",
def->name);
>> + if (def->description)
>> + virBufferVSprintf(&buf, "
<description>%s</description>\n",
>> + def->description);
>> + virBufferVSprintf(&buf, " <state>%s</state>\n",
>> + virDomainStateTypeToString(def->state));
>> + if (def->parent) {
>> + virBufferAddLit(&buf, " <parent>\n");
>> + virBufferVSprintf(&buf, "
<name>%s</name>\n", def->parent);
>> + virBufferAddLit(&buf, " </parent>\n");
>> + }
>> + localtime_r(&def->creationTime, &time_info);
>> + strftime(timestr, sizeof(timestr), "%F_%T", &time_info);
>
> Again, you handle the time in local time. You could use %F_%T%z to
> include the local time zone and then handle that in the parsing
> function to get a correct UTC time back.
>
> Maybe a better solution is to just store the Unix time in seconds in
> UTC (as returned by the time() function) in the XML and let
> applications do the conversion into a more human readable format and
> take care of timezone stuff. This way we get rid of the timezone
> problem at the libvirt level at all.
Well, I was sort of shooting for that by defining the field to be UTC
(which I obviously mis-implemented with localtime_r). It would be nice
to have a human-readable string in the XML, though, which is why I didn't
just use seconds since the Epoch. Can I just use gmtime_r()
here to get the time in UTC, and then strptime and mktime above will do the right thing?
I think the problem is that struct tm doesn't contain a timezone
filed. Therefore, strptime parses a time that is not fixed in a
timezone. So even If we would use a time format like
2010-04-02_12:30:58+0200
it won't help with strptime. strptime knows %z for timezone part, but
just ignores it.
mktime operates in local time. The man page says:
"The mktime() function converts a broken-down time structure,
expressed as local time, to calendar time representation"
So gmtime_r doesn't help.
I think we could use seconds since the Epoch from time() as the actual
value, and attach a human readable version (from strftime with "%a, %d
%b %Y %H:%M:%S %z" in RFC 2822 format) in a comment like this:
<creationTime>1270221648</creationTime><!-- Fri, 02 Apr 2010
17:20:48 +0200 -->
That way we don't need to deal with the subtle timezone stuff as
seconds since the Epoch is in UTC and we have a human readable version
in virsh snapshot-dumpxml for example.
Applications then can take the seconds since the Epoch value and
format it as they like to local time or what ever they prefer.
Matthias