#23 XMLDoc_parse_buffer_DOM bug

[Ermanno]

Hi Matthieu,
I'm using sxmlc lib to read some xml. I receive an error parsing the xml attached. The call chain is:

XMLDoc_parse_buffer_DOM(.. , .. , &doc);
XMLDoc_root(&doc);

The problem is that when XMLDoc_parse_buffer_DOM fails, it calls XMLDoc_free(&doc) and, so, when I call XMLDoc_root it fails as doc is not initialized.

The error is returned by this last function, and is "simple:65: An error was found (UNEXPECTED_TAG_END), loading aborted..."

This error is identified in function _parse_data_SAX into if (txt_end == NULL) { } body.

After testing I've notice that the problem is the '+' char at the end of tag <msginerr>. If I remove it all works. So if there is a '+' char at the end of the value field of a tag, _parse_data_SAX fails.</msginerr>

Thanks for support,

Ermanno

[Matthieu Labas]

Hi Ermanno,
Thank you for your report and the XML file. I will have a look asap.
Matthieu

[Ermanno]

Update: also if you replace the last '+' with '=' it not works.

[Matthieu Labas]

Hi Ermanno,

The following code I am using works fine:

FILE* f = fopen("/tmp/TmRejection.xml", "r");
fseek(f, 0, SEEK_END);
long n = ftell(f);
fseek(f, 0, SEEK_SET);
char* buf = malloc(n+1);
fread(buf, 1, n, f);
buf[n] = 0; // <<<---
fclose(f);
XMLDoc doc;
XMLDoc_init(&doc);
XMLDoc_parse_buffer_DOM(buf, "simple", &doc);
XMLNode* root = XMLDoc_root(&doc);
XMLNode_print(root, stdout, NULL, NULL, 0, 0, 0);

There is indeed a problem when the buffer is not NULL-terminated (if you remove the line marked with <<<---). I will correct it by adding a len parameter.

Could you paste the code you use that shows the error?

Thanks,
Matthieu

[Matthieu Labas]

(which means I suspect the buffer you're parsing is just enough to contain the whole XML and is not null-terminated. Removing a character somewhere will make room for the null termination and make things work)

[Ermanno]

Hi Matthieu,
yesterday evening I detect the problem by debugging. I solve it by using a calloc instead of malloc, so that the buffer is already well terminated. So we are reaching the same problem.

Thanks,
Ermanno

[Matthieu Labas]

v4.2.8 delivered.