[Zlib-devel] known bugs in distance-checking code?

[Greg Roelofs]

>> (1) When the zlib strategy is Z_RLE, are the results dependent on the
>>     compression level (as with default and Z_FILTERED) or independent
>>     (as with Z_HUFFMAN_ONLY)?

> Independent.

OK, that's close to what I'd surmised based on experimentation.  Close,
but not quite identical:  levels 1-3 and 4-9 differ from each other.

   IDAT length with method 137 (fm 0 zl 3 zs 3) =   241049
   IDAT length with method 138 (fm 1 zl 3 zs 3) =   329234
   IDAT length with method 139 (fm 2 zl 3 zs 3) =   195370
   IDAT length with method 140 (fm 3 zl 3 zs 3) =   347977
   IDAT length with method 141 (fm 4 zl 3 zs 3) =   218169 
   IDAT length with method 142 (fm 5 zl 3 zs 3) =   196346

   IDAT length with method 143 (fm 0 zl 4 zs 3) =   240994
   IDAT length with method 144 (fm 1 zl 4 zs 3) =   329163
   IDAT length with method 145 (fm 2 zl 4 zs 3) =   195311
   IDAT length with method 146 (fm 3 zl 4 zs 3) =   347920
   IDAT length with method 147 (fm 4 zl 4 zs 3) =   218107
   IDAT length with method 148 (fm 5 zl 4 zs 3) =   196289

I assume there's some tiny factor in there that makes a 55- to 71-byte
difference some of the time.  (I'm not too concerned as long as there
isn't some other tiny factor that changes things within either levels
1-3 or 4-9.)

Glenn, I'll be sending an update your way shortly.  I've been busy. ;-)

>> (2) Mark, are your pngdat and infgen utilities generally available?

> No.  Oh, wait, if you just received this email, then yes.  They are 
> attached.

Elegantly tiny.  Many thanks!

Greg