Re: [Serusers] [Serdev] [Tracker] Reopened: (SER-263) Using @select=~ "regexp" causes segmentation fault
Samuel,
yes it is related to. The resolution of SER-263 was wrong and every test like if (@select=="") has returned false, even the return value of the select was empty string.
But, there is another problem I see in the config (including the example script)...
There is difference between the (@select) test and (@select!="") one. The latter returns false, if the select framework returns N/A value (e.g. non present header/parameter).
Intended behaviour was: @X is @route[1].params.lr in this example (as the tag must have value if present rfc3261 19.3)
lr parameter test result not present @X false !@X true @X=="" FALSE* @X!="" FALSE*
lr @X true !@X false @X=="" true @X!="" false
lr=123 @X true !@X false @X=="" false @X!="" true
As I see the from the example ser.cfg, the tests are done in the (@to.tag=="") way, which won't work if the to.tag would return either N/A or the tag value. But it seems it returns the empty string even if not present, because it was previously working.
So I ask the community (especially TB potential members) do we want to take care about the select N/A value differently, or shall we make it equal to empty string?
Note, that if the answer is yes, it will be very hard to differentiate between parameter not present and parameter without value - well the lr parameter is handled internally, but what others?
If the anwser is no, keep special N/A value handling, the select must be reviewed together with the ser.cfg, because (@to.tag=="") will be false for every tag (because it must have value).
Michal
On Čt, 2007-06-07 at 13:56 +0200, samuel wrote:
Michal,
I've just post a question to users' list about not properly detecting @to.tag=="".
Is this bug related to it??
Thanks, Samuel.
2007/6/7, Michal Matyska (JIRA) < tracker@iptel.org>: [ http://tracker.iptel.org/browse/SER-263?page=all ]
Michal Matyska reopened SER-263: -------------------------------- The way it is now resolved is wrong. If the select returns empty string the script test if (@select=="") is false. > Using @select=~ "regexp" causes segmentation fault > -------------------------------------------------- > > Key: SER-263 > URL: http://tracker.iptel.org/browse/SER-263 > Project: SER > Issue Type: Bug > Components: core, Selects > Affects Versions: 2.0, Ipteldorf > Environment: CVS head > Reporter: Michal Matyska > Assigned To: Michal Matyska > Priority: Minor > Fix For: 2.0, Ipteldorf > > > If the result of the select is empty string or it is STR_STATIC_INIT ( e.g. @uri.type) the prepatation of left operand (putting \0 behind the string) causes segmentation fault . > Program received signal SIGSEGV, Segmentation fault. > comp_str (op=11, left=0xbfad27a4, rtype=5, r=0x81932b0, msg=0x81a1034) at route.c:668 > 668 left->s[left->len]='\0'; -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://tracker.iptel.org/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira _______________________________________________ Serdev mailing list Serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
Serdev mailing list Serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
IMHO,
I wouldn't "equalize" non-existent parameter with empty string. There are header parameters (maddr,rport,isfocus) which may not have value and someone (who knows?) might need to check for them. I would just not forbid this option from the config file...
I would just use the @ and !@ uses for detecting presence or absence of attributes and letting the == to check for the value once you know it exists. It "complicates" a little bit config file but makes it clearer and I think easier to debug.
So, for me it's fine changing example config files and documenting it.
my 0.02, Samuel
2007/6/8, Michal Matyska michal@iptel.org:
Samuel,
yes it is related to. The resolution of SER-263 was wrong and every test like if (@select=="") has returned false, even the return value of the select was empty string.
But, there is another problem I see in the config (including the example script)...
There is difference between the (@select) test and (@select!="") one. The latter returns false, if the select framework returns N/A value (e.g. non present header/parameter).
Intended behaviour was: @X is @route[1].params.lr in this example (as the tag must have value if present rfc3261 19.3)
lr parameter test result not present @X false !@X true @X=="" FALSE* @X!="" FALSE*
lr @X true !@X false @X=="" true @X!="" false
lr=123 @X true !@X false @X=="" false @X!="" true
As I see the from the example ser.cfg, the tests are done in the (@to.tag=="") way, which won't work if the to.tag would return either N/A or the tag value. But it seems it returns the empty string even if not present, because it was previously working.
So I ask the community (especially TB potential members) do we want to take care about the select N/A value differently, or shall we make it equal to empty string?
Note, that if the answer is yes, it will be very hard to differentiate between parameter not present and parameter without value - well the lr parameter is handled internally, but what others?
If the anwser is no, keep special N/A value handling, the select must be reviewed together with the ser.cfg, because (@to.tag=="") will be false for every tag (because it must have value).
Michal
On Čt, 2007-06-07 at 13:56 +0200, samuel wrote:
Michal,
I've just post a question to users' list about not properly detecting @to.tag=="".
Is this bug related to it??
Thanks, Samuel.
2007/6/7, Michal Matyska (JIRA) < tracker@iptel.org>: [ http://tracker.iptel.org/browse/SER-263?page=all ]
Michal Matyska reopened SER-263: -------------------------------- The way it is now resolved is wrong. If the select returns empty string the script test if (@select=="") is false. > Using @select=~ "regexp" causes segmentation fault > -------------------------------------------------- > > Key: SER-263 > URL: http://tracker.iptel.org/browse/SER-263 > Project: SER > Issue Type: Bug > Components: core, Selects > Affects Versions: 2.0, Ipteldorf > Environment: CVS head > Reporter: Michal Matyska > Assigned To: Michal Matyska > Priority: Minor > Fix For: 2.0, Ipteldorf > > > If the result of the select is empty string or it is STR_STATIC_INIT ( e.g. @uri.type) the prepatation of left operand (putting \0 behind the string) causes segmentation fault . > Program received signal SIGSEGV, Segmentation fault. > comp_str (op=11, left=0xbfad27a4, rtype=5, r=0x81932b0, msg=0x81a1034) at route.c:668 > 668 left->s[left->len]='\0'; -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://tracker.iptel.org/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira _______________________________________________ Serdev mailing list Serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
Serdev mailing list Serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev
Michal Matyska wrote:
Samuel,
yes it is related to. The resolution of SER-263 was wrong and every test like if (@select=="") has returned false, even the return value of the select was empty string.
But, there is another problem I see in the config (including the example script)...
There is difference between the (@select) test and (@select!="") one. The latter returns false, if the select framework returns N/A value (e.g. non present header/parameter).
Intended behaviour was: @X is @route[1].params.lr in this example (as the tag must have value if present rfc3261 19.3)
lr parameter test result not present @X false !@X true @X=="" FALSE* @X!="" FALSE*
lr @X true !@X false @X=="" true @X!="" false
Shouldn't @X=="" return false in this case? (the parameter has no value, not even empty one). To make this true the parameter should be lr="" imho.
IMHO if you specify a value to test in the expression then the parameter must have a value for the test to succeed. If you do not specify any value in the expression then the presence of the parameter should be sufficient for the test to succeed.
lr=123 @X true !@X false @X=="" false @X!="" true
The rest seems fine to me.
As I see the from the example ser.cfg, the tests are done in the (@to.tag=="") way, which won't work if the to.tag would return either N/A or the tag value. But it seems it returns the empty string even if not present, because it was previously working.
As far as I can remember it used to work the way you describe (at least I was using @route[1].params.lr in my configs to test for the presence of the lr parameter some time ago.
So I ask the community (especially TB potential members) do we want to take care about the select N/A value differently, or shall we make it equal to empty string?
No, we should not make it equal. We need to have a mechanism to test for the presence of a parameter regardless its value and yet we need to be able to test if a parameter is set to an empty string.
Note, that if the answer is yes, it will be very hard to differentiate between parameter not present and parameter without value - well the lr parameter is handled internally, but what others?
In digest authentication this could be important. There the absence of a parameter could change the way hash computations are done. The presence of the same parameter, albeit with empty value has different meaning.
If the anwser is no, keep special N/A value handling, the select must be reviewed together with the ser.cfg, because (@to.tag=="") will be false for every tag (because it must have value).
Jan.
Jan Janak wrote:
Shouldn't @X=="" return false in this case? (the parameter has no value, not even empty one). To make this true the parameter should be lr="" imho.
IMHO if you specify a value to test in the expression then the parameter must have a value for the test to succeed. If you do not
Errata: then the parameter must have a value and the value must be equal for the test to succeed.
specify any value in the expression then the presence of the parameter should be sufficient for the test to succeed.
Jan.
-
Jan Janak -
Michal Matyska -
samuel