BUG: CSA reader

[reddigari]

nibabel/nibabel/nicom/csareader.py

Lines 121 to 132 in b46efa3

	# CSA1 specific length modifier
	if tag_no == 1:
	tag0_n_items = n_items
	if n_items > MAX_CSA_ITEMS:
	raise CSAReadError('Expected <= {0} tags, got {1}'.format(
	MAX_CSA_ITEMS, n_items))
	items = []
	for item_no in range(n_items):
	x0, x1, x2, x3 = up_str.unpack('4i')
	ptr = up_str.ptr
	if hdr_type == 1: # CSA1 - odd length calculation
	item_len = x0 - tag0_n_items

I think the definition of tag0_n_items might be a incorrect. The docs (that describe the SPM implementation) say to use the n_items of the first tag, which here corresponds to the condition tag_no == 0, not 1. (Feels like a slip while porting 1-indexed MATLAB code). This would make the value calculated on L132 wrong for every item in every tag.

There are no test files with a CSA1 header, and I don't have any, so I can't confirm this.

[smeisler]

We encountered an error that tag0_n_items was referenced before assignment. Presumably because the tag_no was not 1 (tag0_n_items never defined) but hdr_type was 1 (tag0_n_items used in item_len calculation).