Invalid HTTP responses for custom languages are cached indefinitely

[andreialecu]

Describe the bug

There's an issue in the logic at:

tesseract.js/src/worker-script/index.js

Lines 101 to 104 in adcb5b8

	if (path !== null) {
	const resp = await (isWebWorker ? fetch : adapter.fetch)(`${path}/${lang}.traineddata${gzip ? '.gz' : ''}`);
	data = await resp.arrayBuffer();
	} else {

It doesn't check if the response is successful, which results in a bad response being cached indefinitely. This can happen if the server is down or there's a temporary 404 error, resulting in clients being broken until the IndexedDB is cleared.

This results in the following error being logged on all subsequent usages even if the server error was resolved:

tesseract-core.wasm.js:9 Error opening data file ./mrza.traineddata
tesseract-core.wasm.js:9 Please make sure the TESSDATA_PREFIX environment variable is set to your "tessdata" directory.
tesseract-core.wasm.js:9 Failed loading language 'mrza'
tesseract-core.wasm.js:9 Tesseract couldn't load any languages!

To Reproduce

Use an invalid language and notice that the response is cached in IndexedDb.

Expected behavior

An error should be thrown.

Screenshots

In the screenshot below, mrza is a file that didn't exist, notice the length of the Unit8Array. It's a 404 response.

Additional context

I'll open a PR with a fix shortly.

[Balearica]

Thanks for explaining why this happens--I had the same issue. Hope your fix can get merged soon.