Build system corrupting compressed layer builds by inserting null characters

[peller]

When doing a build, the compression step is inserting a null character [0x00] whenever it encounters x00 [0x5C 0x78 0x30 0x30]. Seems like IE doesn't like seeing a null in the middle of a stream. Probably best to have a special case and leave x00 alone?

In #5017, the "x00" reference was removed from our source tree. This ticket is to fix ShrinkSafe?.

See also #6628

[alex]

I think we only support UTF-8 inputs no matter what. James, are we set up to handle anything like this?

[jburke]

Yeah, I think alex checked in something into the build scripts a little while back that made sure we did utf-8 through the build system. Does that mean we can close this ticket? I do not know if that sequence was particular to a utf-8-related sequence?

[peller]

No, sorry, it was the null JS escape (literally 'x00') that ShrinkSafe? was barfing on, not an ascii=0 in the input stream. It's not common, but it needs to be fixed regardless of the encoding.

[peller]

correction, ShrinkSafe? choke on it, IE did. ShrinkSafe? output a null char, which is a single null byte you're ASCII or UTF-8. The easy solution, iirc, was just to leave x00 alone and not try to turn it into a null char.

[peller]

wait for merge in #7127 if possible

[peller]

fixed in [15594]