Hitomi.la URL Determination Bug

[earfluffy]

So, I'll start by saying that I don't know if this is a problem specific to the gallery I tried to download, something that affects a significant portion of the website, or some weird interaction with Windows again.

In any case, what happened was I attempted to download the gallery found here (NSFW), but got error messages that looked like so:

C:\Users\earfl>gallery-dl -v https://hitomi.la/galleries/1036181.html
gallery-dl: Version 1.6.1
gallery-dl: Python 3.7.1 - Windows-10-10.0.17134-SP0
gallery-dl: requests 2.21.0 - urllib3 1.24.1
gallery-dl: Starting DownloadJob for 'https://hitomi.la/galleries/1036181.html'
hitomi: Using HitomiGalleryExtractor for 'https://hitomi.la/galleries/1036181.html'
urllib3.connectionpool: Starting new HTTPS connection (1): hitomi.la:443
urllib3.connectionpool: https://hitomi.la:443 "GET /galleries/1036181.html HTTP/1.1" 200 2924
urllib3.connectionpool: Starting new HTTPS connection (1): ba.hitomi.la:443
urllib3.connectionpool: https://ba.hitomi.la:443 "GET /galleries/1036181/000.png HTTP/1.1" 403 None
downloader.http: 403 Client Error: Forbidden for url: https://ba.hitomi.la/galleries/1036181/000.png
download: Failed to download hitomi_1036181_001_000.png
urllib3.connectionpool: Starting new HTTPS connection (2): ba.hitomi.la:443
urllib3.connectionpool: https://ba.hitomi.la:443 "GET /galleries/1036181/001.png HTTP/1.1" 403 None
downloader.http: 403 Client Error: Forbidden for url: https://ba.hitomi.la/galleries/1036181/001.png
download: Failed to download hitomi_1036181_002_001.png
urllib3.connectionpool: Starting new HTTPS connection (3): ba.hitomi.la:443
urllib3.connectionpool: https://ba.hitomi.la:443 "GET /galleries/1036181/002.png HTTP/1.1" 403 None
downloader.http: 403 Client Error: Forbidden for url: https://ba.hitomi.la/galleries/1036181/002.png
download: Failed to download hitomi_1036181_003_002.png

     [And so on]
     ...

I went and checked the gallery to see what the URL should have been, and found that it was aa.hitomi.la instead of ba.hitomi.la. I knew it wasn't a config issue, since I don't have any config options for hitomi.la, so I went to look at the extractor file for hitomi.la.

Short story short, I "fixed" the problem by manually changing the part of the extractor file that determines the image URL, i.e., these two lines:

subdomain = chr(97 + self.gid % 2) + "a"
        base = "https://" + subdomain + ".hitomi.la/galleries/"

into this:

subdomain = "a" + "a"
        base = "https://" + subdomain + ".hitomi.la/galleries/"

and then changing it back when I was done. Since this is obviously not the best way to do things, and I don't really understand how the value of self.gid is determined, I figured I ought to report the bug and see if we could figure out the problem. I should note that the other gallery I was interested in downloading worked just fine.

[Hrxn]

It seems like you already found the culprit, good.
Something like this is expected in case of a recent site change, so I'd guess that's what happened here.

[mikf]

Hitomi sets the URLs for displayed images in a few Javascript functions. The important ones can be found at https://ltn.hitomi.la/common.js.

chr(97 + self.gid % 2) is basically what subdomain_from_galleryid(g) does, with number_of_frontends being hard coded to 2 and self.gid representing the gallery ID g (1036181 in this case). You either get aa. or ba. as subdomain, depending on the value of g / self.gid.

What I didn't consider when I last looked at this is that g is not the whole gallery ID, but just the last digit of it, and it gets changed to 0 if it originally was 1:

        if (g === 1) {
                g = 0;
        }

So for any gallery whose ID ends in 1, gallery-dl currently uses the wrong subdomain.