WebKit: in-memory FileReader failover, invalid Content-Range header. about upchunk

[theazgra]

@cjpillsbury I will tag you directly, hope that is ok.

After looking at it more, the problem seems to be with nextChunkRangeStart property. Which is not reset correctly after fallback is applied.

nextChunkRangeStart is updated in successfulChunkUploadCb, then error occurs and readableStreamErrorCallback is called, nextChunkRangeStart is not reset-ed to 0.

This leads to nextChunkRangeStart being different in ChunkedFileIterable instance and UpChunk instance

ResettingnextChunkRangeStart in readableStreamErrorCallback seems to fix the problem.

if (options.useLargeFileWorkaround) {
      const readableStreamErrorCallback = (event: CustomEvent) => {
        // In this case, assume the error is a result of file reading via ReadableStream.
        // Retry using ChunkedFileIterable, which reads the file into memory instead
        // of a stream.
        if (this.chunkedIterable.error) {
          this.nextChunkRangeStart = 0; // <-- reset nextChunkRangeStart
          console.warn(
            `Unable to read file of size ${this.file.size} bytes via a ReadableStream. Falling back to in-memory FileReader!`
          );
          event.stopImmediatePropagation();

          // Re-set everything up with the fallback iterable and corresponding
          // iterator
          this.chunkedIterable = new ChunkedFileIterable(this.file, {
            ...options,
            defaultChunkSize: options.chunkSize,
          });
          this.chunkedIterator = this.chunkedIterable[Symbol.asyncIterator]();
          this.getEndpoint().then(() => {
            this.sendChunks();
          });
          this.off('error', readableStreamErrorCallback);
        }
      };
      this.on('error', readableStreamErrorCallback);
    }

[cjpillsbury]

Hey @theazgra sorry for the delayed response! That looks right to me and all makes sense. Feel free to issue a pull request if you wanted to contribute!