C: preprocessor branches truncated to first line

[randomdude]

Hi all. I'm seeing some unexpected behavior from CodeQL when I process C preprocessor branches that span multiple lines.

For example, if I scan the following C file:

#if defined(FOO) || \
     defined(BAR)
#endif

with the following query:

from PreprocessorBranchDirective pbd select pbd, pbd.toString()

I see the output:

"foo","foo","error","#if defined(FOO) || \","/a.c","1","1","1","21"
"foo","foo","error","#endif","/a.c","3","1","3","6"

I expected to see the entire multiline preprocessor condition in the output of the .toString(), rather than just the first line - but when I went to file this bug, I realised this is not specified anywhere as correct behavior (as far as I can see), and so this may be expected behavior. However, if we ignore that maybe-issue, there is still a second issue - the location information appears to be incorrect, stating that the preprocessor condition ends at line 1 column 21, while it actually ends on line 2.

Finally - it may be that the CSV output is malformed, as the terminating quote is prefixed by a backslash.

Hope this is helpful! For your extra context, I found this while scanning the Linux kernel, so it seems this is a common code pattern, as opposed to an infrequently-seen corner case.

[jketema]

Hi @randomdude,

This is a known issue. I've added a link to this issue to hte internal issue that tracks this.