Hi Vu,

Please see my answers [Lennart]

Thanks
Lennartt

> -----Original Message-----
> From: Vu Minh Nguyen <[email protected]>
> Sent: den 19 juni 2018 07:20
> To: Lennart Lund <[email protected]>; Gary Lee
> <[email protected]>; Hans Nordebäck
> <[email protected]>
> Cc: [email protected]
> Subject: RE: [devel] [PATCH 1/1] imm: fix failure to import file containing
> existing long dn object [#2874]
> 
> Hi Lennart,
> 
> See my comments inline, started with [Vu].
> 
> Regards, Vu
> 
> > -----Original Message-----
> > From: Lennart Lund <[email protected]>
> > Sent: Monday, June 18, 2018 6:02 PM
> > To: Vu Minh Nguyen <[email protected]>; Gary Lee
> > <[email protected]>; Hans Nordebäck
> > <[email protected]>
> > Cc: [email protected]
> > Subject: RE: [devel] [PATCH 1/1] imm: fix failure to import file
> containing
> > existing long dn object [#2874]
> >
> > Hi Vu,
> >
> > Your solution will probably work but is a bit problematic. The actual
> problem
> > is that state is an in-parameter and in this case state->objectName is
> changed
> > by the createImmObject() function and this change is without any meaning
> in
> > the place where createImmObject() is called. Furthermore it seems as if
> the
> > only reason for changing is a trace, TRACE_8("OBJECT NAME: %s", state-
> > >objectName);
> [Vu] The change is there to separate the RDN and the parent DN from
> `state->objectName`.
> The output of this separation is for ` saImmOmCcbObjectCreate_2`.
[Lennart] What's important here is that input data is not changed. It is ok to 
do whatever manipulation of the data that's needed but is must be done using 
local variables, it must not affect the input variables (they should be const)
> 
> > I think you should fix this in an "opposite" way. Instead of creating a
> string
> > variable to preserve the scope you should remove the code that
> manipulates
> > state->objectName, the variable is not owned by createImmObject() so it
> > should never change its value. Instead copy the content of state-
> > >objectName to a temporary string and then manipulate that string in
> order
> > to create the wanted text for the trace.
> [Vu] This is a good idea. I will send out the V2 patch.
> 
> >
> > Should also be fixed (maybe later?):
> > The loadImmXML() is where state is created and initialized. This means
> that
> > this function is the "owner" of the ParserState structure. It also means
> that
> > ParserState (state) shall be within scope as long as this function is in
> scope. If
> > changes are done to variables in the structure by some function it must be
> > clearly stated that this is the case (then state is not an in parameter
> but an
> > in/out parameter) also if state is supposed to be an in-parameter it
> should be
> [Vu] immcfg uses libxml to parse XML file, and ParserState is user data
> which is updated time to time during processing XML file.
> If having any error during that time, following changes/call should be made
> at least:
>     stopParser(state);
>     state->parsingStatus = 1;
> 
> Therefore, ParserState should be an in-out parameter.
[Lennart] It is problematic to use a structure with mixed content as input to a 
function especially if the structure contains a lot of variables and some of 
them is not even related to the function. But as a general rule no input data 
should be manipulated. If the function is supposed to change any variables in 
the structure it should be stated in the function header what variables that is 
considered to be output variables and it must be guaranteed that no other 
variables are changed.
> 
> > const. The structure itself should have a constructor for initializing its
> > variables (is now done in the loadImmXML() function) and a destructor for
> > cleaning up. Now freeing variables is done is several functions!!
> [Vu] I understand your intention, but it could be a lot of works to refactor
> the code using RAII for ParserState.
> Some ParserState data are allocated in this xmlSAXHandler's handler, but
> should be freed in another. Not sure the code can be implemented in the
> way
> as your suggestion.
[Lennart] Yes it may be some work and that's why I suggest that it may be done 
later especially if this fix is urgent. However it does not seems to be that 
much work. An instance of the structure named 'state' is created inside a 
function which means that when this function returns 'state' will go out of 
scope. This function also initialize the 'state' variables. Before 'state' goes 
out of scope all memory that may have been allocated and has pointers in the 
structure must be freed. This seems as the perfect job for a constructor an 
destructor. If some variables are pointers for allocated memory that has to be 
freed several times while 'state' is still in scope cleanup methods can be 
added. However I have not looked deep into this so there may be some problems 
that I am not aware of.
> 
> >
> > Thanks
> > Lennart
> >
> > > -----Original Message-----
> > > From: Vu Minh Nguyen <[email protected]>
> > > Sent: den 11 juni 2018 11:51
> > > To: Gary Lee <[email protected]>; Hans Nordebäck
> > > <[email protected]>
> > > Cc: [email protected]
> > > Subject: [devel] [PATCH 1/1] imm: fix failure to import file containing
> > existing
> > > long dn object [#2874]
> > >
> > > We copied the original object name `state->objectName` using the
> function
> > > osaf_extended_name_lend(state->objectName, &objectName).
> > >
> > > In case of long dn object, objectName contained a reference to the
> string
> > > which was stored in `state->objectName`, any change to `state-
> > > >objectName`,
> > > in this case it was the code line `state->objectName[DNlen] = '\0',
> > > would impact the backup `objectName`.
> > >
> > > As the `objectName` had been changed unintentionally to non-existing
> > > object dn,
> > > `saImmOmAccessorGet_2` returned `SA_AIS_ERR_NOT_EXIST`
> > > unexpectedly.
> > >
> > > To fix this bug, using C++ string to save a copy of the original object
> > > `state->objectName` string. Do so, the string value is not changed when
> > > we play with `state->objectName`.
> > > ---
> > >  src/imm/tools/imm_import.cc | 8 +++++---
> > >  1 file changed, 5 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/src/imm/tools/imm_import.cc
> b/src/imm/tools/imm_import.cc
> > > index 5397c67..d4f7c08 100644
> > > --- a/src/imm/tools/imm_import.cc
> > > +++ b/src/imm/tools/imm_import.cc
> > > @@ -633,7 +633,7 @@ static void createImmObject(ParserState *state) {
> > >    SaAisErrorT errorCode = SA_AIS_OK;
> > >    int i = 0;
> > >    size_t DNlen;
> > > -  SaNameT objectName;
> > > +  std::string objectName;
> > >    std::list<SaImmAttrValuesT_2>::iterator it;
> > >
> > >    TRACE_8("CREATE IMM OBJECT %s, %s", state->objectClass, state-
> > > >objectName);
> > > @@ -647,7 +647,7 @@ static void createImmObject(ParserState *state) {
> > >    osaf_extended_name_clear(&parentName);
> > >    if (state->objectName != NULL) {
> > >      char *parent;
> > > -    osaf_extended_name_lend(state->objectName, &objectName);
> > > +    objectName = std::string{state->objectName};
> > >
> > >      /* ',' is the delimeter */
> > >      /* but '\' is the escape character, used for association objects */
> > > @@ -746,6 +746,7 @@ static void createImmObject(ParserState *state) {
> > >        SaImmAttrValuesT_2 **existing_attributes;
> > >        SaImmAttrValuesT_2 *attr;
> > >        const char *existing_className;
> > > +      SaNameT dn;
> > >
> > >        LOG_IN("OBJECT %s already exist, verifying...",
> state->objectName);
> > >
> > > @@ -759,7 +760,8 @@ static void createImmObject(ParserState *state) {
> > >          goto done;
> > >        }
> > >
> > > -      errorCode = immutil_saImmOmAccessorGet_2(accessorHandle,
> > > &objectName,
> > > +      osaf_extended_name_lend(objectName.c_str(), &dn);
> > > +      errorCode = immutil_saImmOmAccessorGet_2(accessorHandle,
> &dn,
> > >                                                 NULL,  // get all
> attributes
> > >                                                 &existing_attributes);
> > >        if (SA_AIS_OK != errorCode) {
> > > --
> > > 1.9.1
> > >
> > >
> > >
> ----------------------------------------------------------------------------
> --
> > > Check out the vibrant tech community on one of the world's most
> > > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> > > _______________________________________________
> > > Opensaf-devel mailing list
> > > [email protected]
> > > https://lists.sourceforge.net/lists/listinfo/opensaf-devel


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensaf-devel

Reply via email to