Most recent build update fails to load cbz's

[xwaxko]

I have tried 3 different cbzs from different series and they all fail to load

I was not experiencing this problem until the most recent build was updated 2 days ago

Issue occured August 6th

June build worked fine prior to August 6th build

[SlothScript]

I'll look into it. Do you have any logs from the console you could provide?

[SlothScript]

Without details I'm not completely sure of what the issue is, but I've found a few possibilities. Additionally, I realized my commit removed error handling, so that should probably be added back.

Anyway, my theory is that your CBZs may be in the ZIP64 format, and when switching to native ZIP processing in my commit, it didn't support ZIP64 anymore.

Detecting if it's in the ZIP64 format shouldn't be too hard, then we can implement ZIP64.

const zip64Locator = eocd >= 20 && d[eocd-20] === 0x50 && d[eocd-19] === 0x4b && d[eocd-18] === 0x06 && d[eocd-17] === 0x07;
const num = d[eocd+10] | (d[eocd+11] << 8);
const cdOff = (d[eocd+16] | (d[eocd+17] << 8) | (d[eocd+18] << 16) | (d[eocd+19] << 24)) >>> 0;
if (zip64Locator || num === 0xFFFF || cdOff === 0xFFFFFFFF) {
	alert("This CBZ uses ZIP64 extensions, which are not supported yet. Please re-zip it as a standard ZIP file.");
	return 0;
}

[xwaxko]

I was using the deployed github page version so Im not sure if I can get a log from that if there is a way please let me know id be willing to get one for you guys

My console on the page atleast doesnt show any issues

Ive been using the pages version for a few months now since March without issues however as stated previously something in the most recent commit may have broken it

my cbzs have been working fine prior to this and ive tried multiple different series i used to read using this and different browsers and devices but some of them that used to work fail to load like in the screenshot provided while some others do load

[xwaxko]

i just checked whether or not this archive was ZIP64 or not like you said and it shows that it isnt

this exact file i used prior to read using your reader and it used to work fine

so im not sure if this is just an issue on my end

[SlothScript]

Can you send the CBZ?

[xwaxko]

https://gofile.io/d/1JM5D9

I put a random assortment of cbzs in here all of these used to work prior for me

[SlothScript]

Okay, I get the same error as you. I'll look into it further.

[SlothScript]

Okay, I've finally identified the bug.

The native zip parser I implemented the browser native DecompressionStream. This uses a TransformStream, which has a writable side and a readable side.

These streams are driven back pressure, which means that the command to push them only finishes once the browser consumes the bytes. The algorithm to read these cannot run when the output queue is full (which happens due to a large write to the queue). Because the output cannot be consumed, it hangs. The hang comes from a circular wait:

1. writer.write(compressed) is called; Chrome starts decompressing, filling the readable side's internal buffer.
2. Once that buffer (a bounded high-water mark, on the order of a megabyte) fills, the transform pauses, as it can't enqueue more output.
3. The transform stops consuming input. The write() promise never resolves.
4. writer.close() never runs, the reader is never attached, the output is never drained.
5. The await chain in the for loop blocks forever; loadBook never returns.

The reason they sometimes work and sometimes don't is because of the file size. The Soul Eater pages (what I downloaded to test) are 300–580 KB decompressed, which goes way past the limits of the queue.

Anyway, the fix is pretty much this:

const drain = (async () => { while (true) { const {value, done} = await reader.read(); if (done) break; chunks.push(value); } })();
await writer.write(compressed);
await writer.close();
await drain;

Reading and writing run in parallel so the queue never gets full (hopefully).

[SlothScript]

PR made, #3

[xwaxko]

That makes sense it seemed like it was essentially going in some sort of loop when I loaded certain cbzs

Thank you I'll be looking forward to using this again!