Tophat gtf_to_fasta ignoring strand

[fanli]

Hi all,

Am I crazy, or does Tophat's gtf_to_fasta utility seem to ignore strand information? Using the hg19 iGenomes release, for example. Transcript NR_024540 is on the minus strand.

585	NR_024540	chr1	-	14361	29370	29370	29370	11	14361,14969,15795,16606,16857,17232,17605,17914,18267,24737,29320,	14829,15038,15947,16765,17055,17368,17742,18061,18366,24891,29370,	0	WASH7P	unk	unk	-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,

Now, looking at the sequence in UCSC (first two lines):

>NR_024540.1
tccggcagagcggaagcggcggcgggagcttccgggagggcggctcgcag
gcaccatgactcctgtgaggatgcagcactccctggcaggtcagacctat

Looking at the sequence from the genes.fa file that gtf_to_fasta outputs:

>1 NR_024540 chr1- 14362-14829,14970-15038,15796-15947,16607-16765,16858-17055,17233-17368,17606-17742,17915-18061,18268-18366,24738-24891,29321-29370
TCCTGCACAGCTAGAGATCCTTTATTAAAAGCACACTGTTGGTTTCTGCTCAGTTCTTTA
TTGATTGGTGTGCCGTTTTCTCTGGAAGCCTCTTAAGAACACAGTGGCGCAGGCTGGGTG
# nope not the same!

If I reverse complement the whole sequence and then look at the first two lines:

TCCGGCAGAGCGGAAGCGGCGGCGGGAGCTTCCGGGAGGGCGGCTCGCAG
GCACCATGACTCCTGTGAGGATGCAGCACTCCCTGGCAGGTCAGACCTAT
# yay this seems to match!

Now, I sort of remember seeing this a while back, but forget what became of it. Anyone have insight? Is this a bug?