search

bazelbuild / rules_closure

Closure rules for Bazel
https://developers.google.com/closure
Apache License 2.0
152 stars 114 forks source link

Binary files are not correctly served via the phantomjs_harness #196

Open applmak opened 7 years ago

applmak commented 7 years ago

There are two issues here, I guess. One is that the harness does not read the file in binary mode. A fix merely passes 'b' to fs.read when we are serving up a binary file.

The other issue is weirder. When serving a sample png file, the file is changed in some weird way. Specifically:

$ ls -l *.png
-rw-r--r--@ 1 applmak  x  2669 Apr  4 08:24 test1x1.bad.png
-rw-r--r--@ 1 applmak  x  2742 Apr  3 20:12 test1x1.png

Even weirder:

$ xxd test1x1.bad.png | head -n 3
00000000: efbf bd50 4e47 0d0a 1a0a 0000 000d 4948  ...PNG........IH
00000010: 4452 0000 00ef bfbd 0000 00ef bfbd 0803  DR..............
00000020: 0000 00ef bfbd efbf bdef bfbd efbf bd00  ................
$ xxd test1x1.png | head -n 3
00000000: 8950 4e47 0d0a 1a0a 0000 000d 4948 4452  .PNG........IHDR
00000010: 0000 0080 0000 0080 0803 0000 00f4 e091  ................
00000020: f900 0002 7350 4c54 45ff ffff 3333 3333  ....sPLTE...3333

which seems to indicate some kind of encoding issue? This problem might be an issue with phantomjs itself.

applmak commented 7 years ago

https://software.hixie.ch/utilities/cgi/unicode-decoder/utf8-decoder tells me that the strange sequence 0xefbfbd is a result of the first byte of the png being incorrectly interpreted as UTF-8 (0x89) which translates to 

U+FFFD  REPLACEMENT CHARACTER
    * used to replace an incoming character whose value is unknown or unrepresentable in Unicode
    * compare the use of 001A as a control character to indicate the substitute function

The second issue is then a problem is phantomjs. I'll file an issue there and link it here.

applmak commented 7 years ago

Filed ariya/phantomjs#14936 for the encoding issue. Still, the harness needs to be updated.