Skip reassignment of the initialization value

[irwir]

This was incorrect: Reversed condition changed to less instead of less or equal because the value still should be PNG_UINT_31_MAX.
Reassignment of the same value is a common kind of copy-paste errors.
Hence static analysis emits warning in this case.
A better way to avoid reassignment is in the updated commit.

[jbowler]

I don't see any reason for such a change. The compiler can work it out too and making the change requires a code review which explains why the change from:

if (a > b)
  spurious assignment;
else if (a <= b)
  required assignment;

To:

if (a < b)
   required assignment;

Is correct. I couldn't work it out; why is the case "a == b" irrelevant?

However, back to the words of the pull request rather than the code; why is this better documentation of what the code does than the original form, which you present as equivalent (ignoring the "a == b" issue)? It seems to me that "if/else" is always more clear ab initio than "if".

[irwir]

Thanks for reviewing.
I changed the code and the description in the first message.

[jbowler]

libpng-ng comment: libpng(ng) shouldn't even contain this code for IDAT There is no point detecting an over-long IDAT unless it exceeds the PNG spec length (indicating a corrupted stream). IDAT buffering needs to be controlled internally to optimize the specific zlib implementation (if known) and the architecture performance; small buffers or large buffers depend on the relative speed of the zlib window copy vs the libpng row handling for large images and for small images the code can avoid the window copy entirely if enough is known about the zlib implementation.

[irwir]

As could be read, at least one more version of 1.6 was planned. So this code might stay for some time, while 1.7 branch did not pan out.

PS. About 5 years ago in one discussion it was suggested to rename libspng to libpng-ng. It did not happen.

[jbowler]

@ctruta: won't fix