Less Crashy gateway. catch stream2 error

[ya7ya]

This catches the following error. which results in a less crashy gateway as mentioned here #804 (comment)

 Error: write after end
    at writeAfterEnd (/media/op/w3/js-ipfs/node_modules/readable-stream/lib/_stream_writable.js:288:12)
    at PassThrough.Writable.write (/media/op/w3/js-ipfs/node_modules/readable-stream/lib/_stream_writable.js:332:20)
    at pull.through (/media/op/w3/js-ipfs/src/http/gateway/resources/gateway.js:134:21)
    at /media/op/w3/js-ipfs/node_modules/pull-stream/throughs/through.js:17:24
    at drain (/media/op/w3/js-ipfs/node_modules/stream-to-pull-stream/index.js:141:18)
    at Stream.<anonymous> (/media/op/w3/js-ipfs/node_modules/stream-to-pull-stream/index.js:150:5)
    at Stream.emit (events.js:159:13)
    at next (/media/op/w3/js-ipfs/node_modules/pull-stream-to-stream/index.js:120:13)
    at /media/op/w3/js-ipfs/node_modules/pull-stream/throughs/map.js:19:9
    at /media/op/w3/js-ipfs/node_modules/pull-traverse/index.js:42:7

License: MIT
Signed-off-by: Yahya ya7yaz@gmail.com

[daviddias]

@victorbjelkholm Jenkins is not happy anymore:

[victorb]

@diasdavid seems to be about LocalStorage (or similar) has reached the maximum amount of storage allowed.

[daviddias]

@victorbjelkholm It is IndexedDB related.

The maximum browser storage space is dynamic — it is based on your hard drive size. The global limit is calculated as 50% of free disk space.

and

here's also another limit called group limit — this is defined as 20% of the global limit, but it has a minimum of 10MB and a maximum of 2GB.

Source: https://developer.mozilla.org/en-US/docs/Web/API/IndexedDB_API/Browser_storage_limits_and_eviction_criteria

[victorb]

I left this comment somewhere else (on the commit itself I think) but can't find it so leaving it here as well just in case.

While I agree that this patch fixes the particular issue with write after end, it doesn't feel like the completely right fix. We have more uncaught errors/promise rejections that are not solved by this. We need a real fix that catches all errors within the gateway + daemon and prints them, while leaving the gateway + daemon still running.