CMOR segfaults when further_info_url is specified as an empty or pure whitespace string about cmor

[mauzey1]

I have identified where in CMOR this segfault is happening. Below is a section of cmor_CV_checkFurtherInfoURL that checks if the URL template has tokens separated by '<' and '>' characters.

cmor/Src/cmor_CV.c

Lines 379 to 385 in 8d45339

	/* -------------------------------------------------------------------- */
	/* If this is a string with no token we have nothing to do. */
	/* -------------------------------------------------------------------- */
	szToken = strtok(szFurtherInfoURLTemplate, "<>");
	if (strcmp(szToken, cmor_current_dataset.furtherinfourl) == 0) {
	return (0);
	}

When passing an emtpy string to strtok, it will assign a null to szToken. This will cause strcmp to segfault when comparing szToken.

This lead to discovering another behavior of this function: Using the further info URL of the current dataset as a template.

cmor/Src/cmor_CV.c

Lines 373 to 377 in 8d45339

	/* -------------------------------------------------------------------- */
	/* Retrieve default Further URL info */
	/* -------------------------------------------------------------------- */
	strncpy(szFurtherInfoURLTemplate, cmor_current_dataset.furtherinfourl,
	CMOR_MAX_STRING);

If you were to have the following value in your user input...

"further_info_url":       "https://furtherinfo.es-doc.org/<mip_era><source_id>"

then the function will treat that as a template to be turned into a URL for comparison.

It will then proceed to compare what is provided by the user input with the URL created from the template.

C Traceback:
In function: 

!!!!!!!!!!!!!!!!!!!!!!!!!
!
! Warning: The further info in attribute does not match the one found in your Control Vocabulary(CV) File. 
! We found "https://furtherinfo.es-doc.org/<mip_era><source_id>" and 
! CV requires "https://furtherinfo.es-doc.org/CMIP6.PCMDI-test-1-0" 
! Check your Control Vocabulary file "Tables/CMIP6_CV.json".
! 
!
!!!!!!!!!!!!!!!!!!!!!!!!!

So any string that begins with "https://furtherinfo.es-doc.org/" will be valid input according to the CV's further_info_url value, but not if they contain any of the template tokens such as the ones in cmor.h

cmor/include/cmor.h

Line 266 in 8d45339

#define CMOR_DEFAULT_FURTHERURL_TEMPLATE "https://furtherinfo.es-doc.org/<mip_era><institution_id><source_id><experiment_id><sub_experiment_id><variant_label>"

[durack1]

@mauzey1 it would be great to catch this segfault, and rather issue an error message suggesting changes required to the user and quit, does that seems reasonable?