Server returns corrupted .gzipped files

[billyc]

In the UI code, FileAPI.downloadFile() returns the correct blobs for any file OTHER THAN a gzipped file.

Gzipped files are returned with extra bytes somehow; they are corrupted and un-gunzippable.

This could likely be some sort of NGINX interaction problem.

[billyc]

Reference matsim-org/viz#95

[billyc]

The file is definitely "double gzipped".

If I gunzip it once as a binary array of bytes, and then gunzip that result a second time and specifically force it to decode that as a string, I can recover the contents.

const gzdata: any = await BlobUtil.blobToArrayBuffer(blob)
const gunzip1 = pako.inflate(gzdata)
const gunzip2 = pako.inflate(gunzip1, { to: 'string' })

console.log(gunzip2)  // <?xml version 1... etc>

[billyc]

Dropwizard defaults to gzipping things on the fly.

• See https://www.dropwizard.io/1.3.8/docs/manual/configuration.html#gzip

Chrome and Firefox are handling this differently: firefox returns a "double-gzipped" blob whereas Chrome is providing a singly-gzipped blob.

By turning off gzip in Dropwizard, the two browsers both end up with a once-gzipped file.

@Janekdererste is going to explore more fine-grained settings for Dropwizard so that both browsers behave properly, and we don't have to double-unzip everything, and text files still get gzipped for efficiency.