_TEST_FILE_SIZE not 10 KiB

[d912e3]

In test mode, only the first 10 KiB of a file are said to be downloaded and its MD5 hash compared to the value specified in the test case.

While writing a new extractor, I naturally tried to compute the hash as follows:

$ curl $URL | head -c 10K | md5sum

Which only gave a hash that wouldn’t match…

Looking up the definition of the _TEST_FILE_SIZE and the use of it in an HTTP Range request revealed that it isn’t 10 KiB (10240 B) that is downloaded but actually 10241 B since Range: bytes=0-10240 requests the bytes from 0 to 10240, inclusively:

$ curl -r 0-10240 $URL | wc -c
10241
$ curl -r 0-10239 $URL | wc -c
10240

I have no idea what’s the best way to fix this. Changing the definition of _TEST_FILE_SIZE would render probably almost every test case invalid. Updating the respective line in the template may be sufficient.

But maybe the hash is not even meant to be computed the way I tried. Maybe you just enter an arbitrary string for the hash and let the test fail and use the expected string instead…

[dstftw]

Test file size is indeed 10241 bytes and this is probably due to historical reasons. As you noticed, changing the _TEST_FILE_SIZE will require checksum recalculation for each test.
Hovewer, there is an idiomatic and expected way to fetch the file of proper size - you just pass --test to youtube-dl along with URL (this results in downloading first 10241 bytes of the media) following by md5sum on the downloaded file.
Thanks for the report.