This avoids lots of unnecessary work when such streams are referred to via
fetch(), and so their bytes aren't subsequently read. This is a large
performance win on some files.
At the initialization of `Lexer_getObj` (in `parser.js`), there's a loop
that skips whitespace and breaks out whenever EOF is encountered.
(https://github.com/mozilla/pdf.js/blob/88ec2bd1a/src/core/parser.js#L586-L599)
Whenever the current character is not a whitespace character,
`ch = this.nextChar();` is used to find the next character
(using `return this.currentChar = this.stream.getByte())`).
The aforementioned `getByte` method retrieves the next byte using
(https://github.com/mozilla/pdf.js/blob/88ec2bd1a/src/core/stream.js#L122-L128)
var pos = this.pos;
while (this.bufferLength <= pos) {
if (this.eof)
return -1;
this.readBlock();
}
return this.buffer[this.pos++];
This piece of code relies on this.eof to detect whether the last character
has been read. When the stream is a `FlateStream`, and the end of the stream
has been reached, then **`this.eof` is not set to `true`**, because this check
is done inside a loop that does not occur when the read block size is zero:
(https://github.com/mozilla/pdf.js/blob/88ec2bd1ac/src/core/stream.js#L511-L517)
for (var n = bufferLength; n < end; ++n) {
if (typeof (b = bytes[bytesPos++]) == 'undefined') {
this.eof = true;
break;
}
buffer[n] = b;
}
This commit fixes the issue by setting this.eof to true whenever the loop is not
going to run (i.e. when bufferLength === end, i.e. blockLen === 0).
