Menu
▾
▴
Re: [devel] [PATCH 1 of 1] imm: Canonicalize attributes presented by OiCcbObjectModifyCallback [#801]
|
From: Neelakanta R. <red...@or...> - 2015-12-18 08:51:40
|
Hi AndersBj/Hung,
The present #801 may be handled for only Canonicalize attributes
presented by OiCcbObjectModifyCallback.
Sending All attributes including NOT writable attributes is debatable
and not backward compatible for appliers also.
A new Enhancement ticket may be opened which talks about sending all
attribute to appliers or for both appliers and OIs.
Thanks,
Neel.
On Thursday 17 December 2015 06:44 PM, Anders Bjornerstedt wrote:
> H iNeel,
> Comments inline.
>
>
>> ----Ursprungligt meddelande----
>> Från : red...@or...
>> Datum : 17-12-2015 - 12:48 (GMT)
>> Till : and...@co..., hun...@de...
>> Kopia : ope...@li...
>> Ämne : Re: [devel] [PATCH 1 of 1] imm: Canonicalize attributes presented by OiCcbObjectModifyCallback [#801]
>>
>> Hi AndersBj,
>>
>> 1. In the previous discussion, the following is the differentiation
>> between OI and applier:
>>
>> snippet from previous discussion:
>>
>> Converning point (c): I would say that how light or heavy an appplier vs OI is is completely up to the developers of these functions.
>> They should preferrably get the same callbacks with the same contents as the OI unless there is a very good reason to break that rule.
>> The reason is the old KISS rule. Not folowing it will result in the average OI/applier developer getting it wrong. As we all know the average
>> developer does not read much if any documentation, unless they really run into problems...
> In this case we have a good reason to very slightly "break the rule".
> By "very slightly" I mean that we are not breaking any rule of ligical equivalence.
> The only "rule" we would be breaking by sending all attributes to appliers (if they have initialized with the latest release)
> is a a silly "rule" that says that the actual set of values at every instance of a cllaback has to be the same.
> That is a much much stromnger rule and I claim it is unnecesarily restrictive.
> It has the disadvantage that we would not be allowed to treat the applier in any way diferently (in the callbacks) despite the fact
> that it is already treated differently. Appliers can not reply to the callbacks and appliers are not supposed to validate.
> We should avoid making things more different and complex than they need to be.
> But we should not be afraid to make some differences if tyhose differences actually simplify the job of the eveloper.
>
>> 2. Going by the real usage of(OI and applier) this (in middleware
>> services AMF ) :
>>
>> The code for active and standby(where applier) has similar code.
>> The only difference is the standby(applier) will not validate.
> So how would the AMF applier have a problem with receiving all the values ?
> Have you actually looked at the AMF code and checked this ?
> I suspect (but I have not checked) that the only problem the AMF applier has is that it can not handle
> receiving values for NOT writable attributes, even when those values are identical to the existing ones.
> That is a minor issue in the AMF applier and is very easy to fix.
> The advantage is that we solve the applier initialization prooblem which is a really hard problem currently with no
> no good solution. The AMF OI will not be impacted as long as it does not increas its API version.
>
>> 3. If the writable attributes are not sent, still applier want to
>> initialize then the applier has to read using searchnext/accessorget .
>> Then, sending all writable attributes may not be useful as intended.
> I think you mean "if the NOT writable attributes are not sent" or ?
> I am actually saying that all attributes including NOT writable attributes could be sent to appliers.
> An applier that initializes with the latest version will only do so because it wants to get what that new version
> provides. It will then know that it is getting the after value of all attributes.
> That would solve the initialization problem for appliers and the applier would know that if the attribute is not writable
> then the received value must be the same as any previous one. It could even assert on such an equality check, except when
> it is initializing.
>
> So in summary.
> I just dont see what the AMFs big hangup is with this. Why the AMF has to prevent a solution for the initialization pproblem for
> ALL appliers for ALL future just becausee iit parhaps has some extra ooverzealos check that can be removed the day the AMF wants
> to uppgrade its OI.
>
> /AndersBj
>
>> Thanks,
>> Neel.
>>
>>
>> On Thursday 17 December 2015 05:27 PM, Anders Bjornerstedt wrote:
>>> Hi
>>>
>>> (having problems getting through to the list).
>>>
>>> I think Hungs's opatch was/is ok. Otherwise i would have commented it.
>>>
>>> Point (1) as I see it should be taken in the sense of logic, not of conctent.
>>> Thus the basic logic for correctly receiving the callback should be the same in both OI anbd applier.
>>> But the content (values actually received) does not necessarily have to be the same.
>>> Specifically, the applier should not do any validation or checking of the values.
>>> The applier should always simply record the values and apply them if/when it receives the apply callback.
>>> The completed callback has no meaning for appliers.
>>>
>>> Point (2) as I believe to drop sending all values to the regular OI.
>>> It was not necessary to drop sendiing all values to apppliers.
>>> Neel if you see some problem with this then please explain what that problem would be.
>>> The only possible problem I could see is if a value appears for an attribute that does not have the writable flag
>>> and the applier checks against tthat. But the applier should not do such checking anyway. It is not the appliers
>>> job to do so. In addition if it really did check the value it would discover that it was the same value as the current one,
>>> i.e. no change logically. So even if the aplier did "validate" (which it should not do) it should still not have a logical problem
>>> with the value.
>>>
>>> The significant *advantage* with getting all values in the applier callback I have already explained.
>>> It solves the initialization pproblem.
>>>
>>> I dont see any point in adding this feature later, if we have it implemented now.
>>> If there is a "problem" in adding this now with the version check in olace, then there will always be a problem and it will
>>> never get added. Converesely If there is a problem now it would only be for an applier that explicitly sets to the new API version
>>> and such an applier has oovverzealous checking of the type reject a replacement with the same value as the current one for
>>> attributes that dont have the writable flag set.
>>>
>>> /AndersBj
>>>
>>>> ----Ursprungligt meddelande----
>>>> Från : hun...@de...
>>>> Datum : 17-12-2015 - 10:37 (GMT)
>>>> Till : red...@or..., zor...@er...
>>>> Kopia : ope...@li...
>>>> Ämne : Re: [devel] [PATCH 1 of 1] imm: Canonicalize attributes presented by OiCcbObjectModifyCallback [#801]
>>>>
>>>> Hi Neel,
>>>>
>>>> I misunderstood the discussion, I will fix that.
>>>>
>>>> In future, do you think we will implement that sending-all-writable-attributes feature?
>>>> If so, I will keep the code for canonicalizing all writable attributes. We just simply don't use it for now.
>>>> Otherwise, I will clean up the unused code.
>>>>
>>>> Thanks,
>>>>
>>>> Hung Nguyen - DEK Technologies
>>>>
>>>>
>>>> --------------------------------------------------------------------------------
>>>> From: Neelakanta Reddy red...@or...
>>>> Sent: Thursday, December 17, 2015 5:02PM
>>>> To: Hung Nguyen, Zoran Milinkovic
>>>> hun...@de..., zor...@er...
>>>> Cc: Opensaf-devel
>>>> ope...@li...
>>>> Subject: Re: [PATCH 1 of 1] imm: Canonicalize attributes presented by OiCcbObjectModifyCallback [#801]
>>>>
>>>>
>>>> Hi Hung,
>>>>
>>>> The following is the final conclusion from the discussion and from
>>>> AndersBj( view point):
>>>>
>>>> 1. Both Appliers and implementer "*must*" receive the same
>>>> modification values for modify callback.
>>>> 2. The "final" proposal below was to drop the notion of sending all
>>>> values.
>>>> 3. The only change the enhancement should consider is, for modify
>>>> callback for appliers and implementers from A.2.17 will have atrrModType
>>>> as SA_IMM_ATTR_VALUES_REPLACE. Only modify attributes must be sent to
>>>> modifycallback. Only afterimage values must be sent through modify callback.
>>>>
>>>> The published patch has :
>>>> a. For applier modify callback all writable attributes are sent (this
>>>> violates point 1 above)
>>>> b. The, PBE must also be compatible with new changes and the PBE
>>>> callback must also be sent with canonicalize attributes.
>>>>
>>>> /Neel.
>>>>
>>>>
>>>> On Monday 14 December 2015 03:02 PM, Hung Nguyen wrote:
>>>>> osaf/services/saf/immsv/immnd/ImmModel.cc | 142 +++++++++++++++++++++++++++++
>>>>> osaf/services/saf/immsv/immnd/ImmModel.hh | 8 +
>>>>> osaf/services/saf/immsv/immnd/immnd_evt.c | 66 +++++++++++-
>>>>> osaf/services/saf/immsv/immnd/immnd_init.h | 2 +
>>>>> 4 files changed, 209 insertions(+), 9 deletions(-)
>>>>>
>>>>>
>>>>> If the modType is SA_IMM_ATTR_VALUES_ADD or SA_IMM_ATTR_VALUES_DELETE,
>>>>> it will be converted to SA_IMM_ATTR_VALUES_REPLACE and the values of the attr-mod will be the values in the after image.
>>>>>
>>>>> This feature is only for OIs with version A.2.17 or later so we don't modify the content of incoming IMMND_EVT_A2ND_OBJ_MODIFY messages.
>>>>> New memory is allocated for canonicalized attribute-modifications instead.
>>>>>
>>>>> OIs and appliers which are initialized as A.2.17 or later MUST support SA_IMM_ATTR_VALUES_REPLACE in their callbacks.
>>>>>
>>>>> diff --git a/osaf/services/saf/immsv/immnd/ImmModel.cc b/osaf/services/saf/immsv/immnd/ImmModel.cc
>>>>> --- a/osaf/services/saf/immsv/immnd/ImmModel.cc
>>>>> +++ b/osaf/services/saf/immsv/immnd/ImmModel.cc
>>>>> @@ -717,6 +717,11 @@ immModel_genSpecialModify(IMMND_CB *cb,
>>>>> genSpecialModify(req);
>>>>> }
>>>>>
>>>>> +struct immsv_attr_mods_list*
>>>>> +immModel_canonicalizeAttrModification(IMMND_CB *cb, const struct ImmsvOmCcbObjectModify *req, bool allWritable, bool* hasLongDns)
>>>>> +{
>>>>> + return ImmModel::instance(&cb->immModel)->canonicalizeAttrModification(req, allWritable, hasLongDns);
>>>>> +}
>>>>>
>>>>> SaUint32T
>>>>> immModel_getLocalAppliersForObj(IMMND_CB *cb,
>>>>> @@ -6996,6 +7001,143 @@ ImmModel::specialApplierTrimCreate(SaUin
>>>>> return attrValues;
>>>>> }
>>>>>
>>>>> +immsv_attr_mods_list*
>>>>> +ImmModel::attrValueToAttrMod(const ObjectInfo* obj, const std::string& attrName,
>>>>> + SaUint32T attrType, bool* hasLongDns)
>>>>> +{
>>>>> + ImmAttrValueMap::const_iterator avi = obj->mAttrValueMap.find(attrName);
>>>>> + osafassert(avi != obj->mAttrValueMap.end());
>>>>> + ImmAttrValue* attrValue = avi->second;
>>>>> +
>>>>> + immsv_attr_mods_list* attrMod = (immsv_attr_mods_list*) calloc(1, sizeof(immsv_attr_mods_list));
>>>>> + osafassert(attrMod);
>>>>> + attrMod->attrModType = SA_IMM_ATTR_VALUES_REPLACE;
>>>>> +
>>>>> + /* attrValue.attrValueType */
>>>>> + attrMod->attrValue.attrValueType = attrType;
>>>>> +
>>>>> + /* attrValue.attrName */
>>>>> + attrMod->attrValue.attrName.size = strlen(attrName.c_str()) + 1;
>>>>> + attrMod->attrValue.attrName.buf = (char*) malloc(attrMod->attrValue.attrName.size);
>>>>> + osafassert(attrMod->attrValue.attrName.buf);
>>>>> + strncpy(attrMod->attrValue.attrName.buf, attrName.c_str(), attrMod->attrValue.attrName.size);
>>>>> +
>>>>> + /* attrValue.attrValuesNumber, attrValue.attrValue and attrValue.attrMoreValues */
>>>>> + attrMod->attrValue.attrValuesNumber = 0;
>>>>> + if (!attrValue->empty()) {
>>>>> + attrValue->copyValueToEdu(&(attrMod->attrValue.attrValue), (SaImmValueTypeT) attrType);
>>>>> + if (attrValue->extraValues()) {
>>>>> + ImmAttrMultiValue* multiVal = (ImmAttrMultiValue *) attrValue;
>>>>> + multiVal->copyExtraValuesToEdu(&(attrMod->attrValue.attrMoreValues), (SaImmValueTypeT) attrType);
>>>>> + }
>>>>> + attrMod->attrValue.attrValuesNumber = 1 + attrValue->extraValues();
>>>>> + } /* else, attrValuesNumber is already set to 0 */
>>>>> +
>>>>> + /* Check for long DN. Will be skipped if hasLongDns is NULL or already 'true' */
>>>>> + if (hasLongDns && !(*hasLongDns) && !attrValue->empty()) {
>>>>> + /* Check head value */
>>>>> + if (strlen(attrValue->getValueC_str()) >= SA_MAX_UNEXTENDED_NAME_LENGTH) {
>>>>> + *hasLongDns = true; /* Found! Skip checking more values */
>>>>> + } else if (attrValue->extraValues()) {
>>>>> + /* Check more values */
>>>>> + ImmAttrMultiValue* multiVal = ((ImmAttrMultiValue *) attrValue)->getNextAttrValue();
>>>>> + while (multiVal) {
>>>>> + if (!multiVal->empty() && strlen(multiVal->getValueC_str()) >= SA_MAX_UNEXTENDED_NAME_LENGTH) {
>>>>> + *hasLongDns = true; /* Found! Stop checking */
>>>>> + break; /* while */
>>>>> + }
>>>>> + multiVal = multiVal->getNextAttrValue();
>>>>> + }
>>>>> + }
>>>>> + }
>>>>> +
>>>>> + return attrMod;
>>>>> +}
>>>>> +
>>>>> +/* This function allocates new memory for canonicalized attribute-modifications.
>>>>> + * The 'attrMods' of input ImmsvOmCcbObjectModify remains untouched.
>>>>> + * Remember to free the memory with immsv_free_attrmods().
>>>>> + *
>>>>> + * 'hasLongDns' is an in/out argument. If it's NULL or 'true', no checking for long DNs will be done.
>>>>> + * 'hasLongDns' will only be returned when canonicalizing all writable attributes.
>>>>> + * In case of canonicalizing only modified attributes, we can use 'hasLongDns' ouput
>>>>> + * from immModel_ccbObjectModify(), no need to check the attributes for long DN again.
>>>>> + */
>>>>> +immsv_attr_mods_list*
>>>>> +ImmModel::canonicalizeAttrModification(const ImmsvOmCcbObjectModify *req, bool allWritable, bool* hasLongDns)
>>>>> +{
>>>>> + TRACE_ENTER();
>>>>> + immsv_attr_mods_list* result = NULL;
>>>>> + std::string objectName;
>>>>> + CcbVector::iterator ci;
>>>>> + CcbInfo* ccb = NULL;
>>>>> + ObjectInfo* afim = NULL;
>>>>> + ObjectMutationMap::iterator omuti;
>>>>> + ObjectMutation* oMut = NULL;
>>>>> +
>>>>> + /* Get object Name */
>>>>> + size_t sz = strnlen(req->objectName.buf, (size_t) req->objectName.size);
>>>>> + objectName.append((const char*) req->objectName.buf, sz);
>>>>> + osafassert(!objectName.empty());
>>>>> +
>>>>> + /* Get ccb info */
>>>>> + ci = std::find_if(sCcbVector.begin(), sCcbVector.end(), CcbIdIs(req->ccbId));
>>>>> + osafassert(ci != sCcbVector.end());
>>>>> + ccb = *ci;
>>>>> +
>>>>> + /* Get object mutation */
>>>>> + omuti = ccb->mMutations.find(objectName);
>>>>> + osafassert(omuti != ccb->mMutations.end());
>>>>> + oMut = omuti->second;
>>>>> +
>>>>> + /* Get after image */
>>>>> + osafassert(oMut->mOpType != IMM_DELETE);
>>>>> + if (oMut->mOpType == IMM_CREATE) { /* Chained operation */
>>>>> + ObjectMap::iterator oi = sObjectMap.find(objectName);
>>>>> + osafassert(oi != sObjectMap.end());
>>>>> + afim = oi->second;
>>>>> + } else if (oMut->mOpType == IMM_MODIFY) {
>>>>> + afim = oMut->mAfterImage;
>>>>> + }
>>>>> +
>>>>> + if (allWritable) {
>>>>> + /* add all writable attributes of class to list */
>>>>> + osafassert(hasLongDns);
>>>>> + ClassInfo* classInfo = afim->mClassInfo;
>>>>> + AttrMap::iterator ai = classInfo->mAttrMap.begin();
>>>>> + for (; ai != classInfo->mAttrMap.end(); ai++) {
>>>>> + if (ai->second->mFlags & SA_IMM_ATTR_WRITABLE) {
>>>>> + immsv_attr_mods_list* attrMod = NULL;
>>>>> + if (((ai->second->mValueType == SA_IMM_ATTR_SANAMET) ||
>>>>> + (ai->second->mValueType == SA_IMM_ATTR_SASTRINGT && (ai->second->mFlags & SA_IMM_ATTR_DN)))) {
>>>>> + attrMod = attrValueToAttrMod(afim, (std::string&) ai->first,
>>>>> + ai->second->mValueType, hasLongDns);
>>>>> + } else {
>>>>> + attrMod = attrValueToAttrMod(afim, (std::string&) ai->first,
>>>>> + ai->second->mValueType, NULL);
>>>>> + }
>>>>> + attrMod->next = result;
>>>>> + result = attrMod;
>>>>> + }
>>>>> + }
>>>>> + } else {
>>>>> + /* add only modified attributes to list */
>>>>> + immsv_attr_mods_list* reqAttrMods = req->attrMods;
>>>>> + for (; reqAttrMods; reqAttrMods = reqAttrMods->next) {
>>>>> + size_t sz = strnlen(reqAttrMods->attrValue.attrName.buf,
>>>>> + (size_t) reqAttrMods->attrValue.attrName.size);
>>>>> + std::string attrName((const char *) reqAttrMods->attrValue.attrName.buf, sz);
>>>>> + immsv_attr_mods_list* attrMod = attrValueToAttrMod(afim, attrName,
>>>>> + reqAttrMods->attrValue.attrValueType, NULL);
>>>>> + attrMod->next = result;
>>>>> + result = attrMod;
>>>>> + }
>>>>> + }
>>>>> +
>>>>> + TRACE_LEAVE();
>>>>> + return result;
>>>>> +}
>>>>> +
>>>>> /**
>>>>> * Creates an object
>>>>> */
>>>>> diff --git a/osaf/services/saf/immsv/immnd/ImmModel.hh b/osaf/services/saf/immsv/immnd/ImmModel.hh
>>>>> --- a/osaf/services/saf/immsv/immnd/ImmModel.hh
>>>>> +++ b/osaf/services/saf/immsv/immnd/ImmModel.hh
>>>>> @@ -239,6 +239,9 @@ public:
>>>>> SaUint32T ccbId,
>>>>> ConnVector& connVector,
>>>>> SaUint32T* appCtnPtr);
>>>>> +
>>>>> + immsv_attr_mods_list* canonicalizeAttrModification(
>>>>> + const struct ImmsvOmCcbObjectModify *req, bool allWritable, bool* hasLongDns);
>>>>>
>>>>> SaAisErrorT ccbObjectModify(
>>>>> const ImmsvOmCcbObjectModify* req,
>>>>> @@ -709,6 +712,11 @@ private:
>>>>> ObjectInfo *obj,
>>>>> const char *noDanglingRef);
>>>>>
>>>>> + immsv_attr_mods_list* attrValueToAttrMod(const ObjectInfo* obj,
>>>>> + const std::string& attrName,
>>>>> + SaUint32T attrType,
>>>>> + bool* hasLongDns);
>>>>> +
>>>>> };
>>>>>
>>>>> #endif
>>>>> diff --git a/osaf/services/saf/immsv/immnd/immnd_evt.c b/osaf/services/saf/immsv/immnd/immnd_evt.c
>>>>> --- a/osaf/services/saf/immsv/immnd/immnd_evt.c
>>>>> +++ b/osaf/services/saf/immsv/immnd/immnd_evt.c
>>>>> @@ -6212,6 +6212,13 @@ static void immnd_evt_proc_object_modify
>>>>> SaNameT objName;
>>>>> osaf_extended_name_clear(&objName);
>>>>> bool hasLongDns=false;
>>>>> + /* These 2 attr-mods lists will only be generated on demand */
>>>>> + IMMSV_ATTR_MODS_LIST* canonicalizedAttrMod = NULL;
>>>>> + IMMSV_ATTR_MODS_LIST* canonicalizedWritableAttr = NULL;
>>>>> + /* Used when canonicalizing all writable attributes */
>>>>> + bool writableAttrHasLongDns = false;
>>>>> +
>>>>> +
>>>>> TRACE_ENTER();
>>>>> #if 0 /*ABT DEBUG PRINTOUTS START */
>>>>> TRACE_2("ABT immnd_evt_proc_object_modify object:%s", evt->info.objModify.objectName.buf);
>>>>> @@ -6243,6 +6250,11 @@ static void immnd_evt_proc_object_modify
>>>>> err = immModel_ccbObjectModify(cb, &(evt->info.objModify), &implConn, &implNodeId,
>>>>> &continuationId, &pbeConn, pbeNodeIdPtr, &objName, &hasLongDns);
>>>>>
>>>>> + /* If 'hasLongDns' is already true, canonicalizedWritableAttr will also contain long DN.
>>>>> + * No need to do any check for long DN in immModel_canonicalizeAttrModification.
>>>>> + */
>>>>> + writableAttrHasLongDns = hasLongDns;
>>>>> +
>>>>> if(pbeNodeIdPtr && pbeConn && err == SA_AIS_OK) {
>>>>> /*The persistent back-end is present and executing at THIS node. */
>>>>> osafassert(cb->mIsCoord);
>>>>> @@ -6317,6 +6329,19 @@ static void immnd_evt_proc_object_modify
>>>>> IMMA_EVT_ND2A_OI_OBJ_MODIFY_LONG_UC : IMMA_EVT_ND2A_OI_OBJ_MODIFY_UC;
>>>>>
>>>>> send_evt.info.imma.info.objModify = evt->info.objModify;
>>>>> +
>>>>> + /* For A.2.17 or later */
>>>>> + if (oi_cl_node->version.minorVersion >= 0x11 &&
>>>>> + oi_cl_node->version.majorVersion == 0x2 &&
>>>>> + oi_cl_node->version.releaseCode == 'A') {
>>>>> + if (!canonicalizedAttrMod) {
>>>>> + canonicalizedAttrMod =
>>>>> + immModel_canonicalizeAttrModification(cb, &(evt->info.objModify),
>>>>> + false, NULL);
>>>>> + }
>>>>> + send_evt.info.imma.info.objModify.attrMods = canonicalizedAttrMod;
>>>>> + }
>>>>> +
>>>>> /* shallow copy into stack alocated structure. */
>>>>>
>>>>> send_evt.info.imma.info.objModify.adminOwnerId = continuationId;
>>>>> @@ -6350,14 +6375,16 @@ static void immnd_evt_proc_object_modify
>>>>> if(arrSize) {
>>>>> memset(&send_evt, '\0', sizeof(IMMSV_EVT));
>>>>> send_evt.type = IMMSV_EVT_TYPE_IMMA;
>>>>> - send_evt.info.imma.type = hasLongDns ? IMMA_EVT_ND2A_OI_OBJ_MODIFY_LONG_UC :
>>>>> - IMMA_EVT_ND2A_OI_OBJ_MODIFY_UC;
>>>>> send_evt.info.imma.info.objModify = evt->info.objModify;
>>>>> send_evt.info.imma.info.objModify.adminOwnerId = 0;
>>>>> /* Re-use the adminOwner member of the ccbModify message to hold the
>>>>> invocation id. In this case, 0 => no reply is expected. */
>>>>>
>>>>> for (; ix < arrSize && err == SA_AIS_OK; ++ix) {
>>>>> + bool isSpecialApplier = false;
>>>>> + send_evt.info.imma.info.objModify.attrMods = evt->info.objModify.attrMods;
>>>>> + send_evt.info.imma.type = hasLongDns ?
>>>>> + IMMA_EVT_ND2A_OI_OBJ_MODIFY_LONG_UC : IMMA_EVT_ND2A_OI_OBJ_MODIFY_UC;
>>>>> implHandle = m_IMMSV_PACK_HANDLE(applConnArr[ix], cb->node_id);
>>>>> send_evt.info.imma.info.objModify.immHandle = implHandle;
>>>>>
>>>>> @@ -6367,14 +6394,36 @@ static void immnd_evt_proc_object_modify
>>>>> If not special applier, then attribute-list will be untouched
>>>>> since the last applier is then a regular applier.
>>>>> */
>>>>> - evt->info.objModify.attrMods = send_evt.info.imma.info.objModify.attrMods =
>>>>> + send_evt.info.imma.info.objModify.attrMods =
>>>>> immModel_specialApplierTrimModify(cb, applConnArr[ix], &(evt->info.objModify));
>>>>>
>>>>> + /* If attrMods of 'send_evt' is different from attrMods of 'evt'
>>>>> + * then we are sending to the special applier. */
>>>>> + if (send_evt.info.imma.info.objModify.attrMods != evt->info.objModify.attrMods) {
>>>>> + isSpecialApplier = true;
>>>>> + evt->info.objModify.attrMods = send_evt.info.imma.info.objModify.attrMods;
>>>>> + }
>>>>> }
>>>>>
>>>>> /*Fetch client node for Applier OI ! */
>>>>> immnd_client_node_get(cb, implHandle, &oi_cl_node);
>>>>> osafassert(oi_cl_node != NULL);
>>>>> +
>>>>> + /* For A.2.17 or later */
>>>>> + if (!isSpecialApplier &&
>>>>> + oi_cl_node->version.minorVersion >= 0x11 &&
>>>>> + oi_cl_node->version.majorVersion == 0x2 &&
>>>>> + oi_cl_node->version.releaseCode == 'A') {
>>>>> + if (!canonicalizedWritableAttr) {
>>>>> + canonicalizedWritableAttr =
>>>>> + immModel_canonicalizeAttrModification(cb, &(evt->info.objModify),
>>>>> + true, &writableAttrHasLongDns);
>>>>> + }
>>>>> + send_evt.info.imma.info.objModify.attrMods = canonicalizedWritableAttr;
>>>>> + send_evt.info.imma.type = writableAttrHasLongDns ?
>>>>> + IMMA_EVT_ND2A_OI_OBJ_MODIFY_LONG_UC : IMMA_EVT_ND2A_OI_OBJ_MODIFY_UC;
>>>>> + }
>>>>> +
>>>>> if (oi_cl_node->mIsStale) {
>>>>> LOG_WA("Applier client went down so modify upcall not sent");
>>>>> continue;
>>>>> @@ -6416,12 +6465,11 @@ static void immnd_evt_proc_object_modify
>>>>> }
>>>>>
>>>>> done:
>>>>> - /*Free the incomming events substructure. */
>>>>> - free(evt->info.objModify.objectName.buf);
>>>>> - evt->info.objModify.objectName.buf = NULL;
>>>>> - evt->info.objModify.objectName.size = 0;
>>>>> - immsv_free_attrmods(evt->info.objModify.attrMods);
>>>>> - evt->info.objModify.attrMods = NULL;
>>>>> + /* Free the canonicalized attr mods */
>>>>> + immsv_free_attrmods(canonicalizedAttrMod);
>>>>> + canonicalizedAttrMod = NULL;
>>>>> + immsv_free_attrmods(canonicalizedWritableAttr);
>>>>> + canonicalizedWritableAttr = NULL;
>>>>> osaf_extended_name_free(&objName);
>>>>> TRACE_LEAVE();
>>>>> }
>>>>> diff --git a/osaf/services/saf/immsv/immnd/immnd_init.h b/osaf/services/saf/immsv/immnd/immnd_init.h
>>>>> --- a/osaf/services/saf/immsv/immnd/immnd_init.h
>>>>> +++ b/osaf/services/saf/immsv/immnd/immnd_init.h
>>>>> @@ -296,6 +296,8 @@ extern "C" {
>>>>>
>>>>> void immModel_genSpecialModify(IMMND_CB *cb, struct ImmsvOmCcbObjectModify *req);
>>>>>
>>>>> + struct immsv_attr_mods_list*
>>>>> + immModel_canonicalizeAttrModification(IMMND_CB *cb, const struct ImmsvOmCcbObjectModify *req, bool allWritable, bool* hasLongDns);
>>>>>
>>>>> SaBoolT immModel_protocol41Allowed(IMMND_CB *cb);
>>>>> SaBoolT immModel_protocol43Allowed(IMMND_CB *cb);
>>>> ------------------------------------------------------------------------------
>>>> _______________________________________________
>>>> Opensaf-devel mailing list
>>>> Ope...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/opensaf-devel
>>>>
>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> Opensaf-devel mailing list
>> Ope...@li...
>> https://lists.sourceforge.net/lists/listinfo/opensaf-devel
>>
|