BUG-FIX 1.6.45 does not build

[jbowler]

This removes use of non-ANSI features in 1.6.45 and 1.6.46.

Signed-off-by: John Bowler [email protected]

[jbowler]

Changes: corrects extent of "int error" in png_XYZ_from_xy and, thereby, corrects the incorrect "declaration after statement" bug that I introduced. Also corrects the indentation and line-width errors that introduction of the new extent generates; stupid GIT.

Corrects seven spurious semicolons. For not-so-good reasons PNG_FIXED_EXPORT and FP_FP_EXPORT provide their own semicolons. I forgot. My bad. Fixed.

Introduces: spurious end-of-line spaces in aclocal.m4. I blame autoreconf, well, I blame git too - git commit -s -a. I don't much care about this latter.

[ctruta]

FYI, I wanted to integrate this today, along with build and CI changes of the kind that you suggested, but some other task popped up. I want to finish it tomorrow, though, and to release libpng-1.6.46 with that.

[jbowler]

I can't see any updates to 1.6.45 yet so a quick release of .46 will help distribution maintainers. Here's my count:

Ubuntu ARM64: libpng16-16t64 1.6.43-5
This was upgraded very recently from 1.6.37 so I assume their upgrades take a while...
Ubuntu ARM32: libpng16-16 1.6.37-3build5
This is an RPi with an ARMv7a and they've pretty much dropped support for those.
Gentoo ~amd64: libpng-1.6.44
There is an upgrade "notice" on the maintainer page, so if you release 1.6.46 now it will probably go in rather than 1.6.45. It's taking longer than I expected but I see no public record of any issues the maintainers have.

I also have more PNGv3 conformance stuff connected to the breakage caused by the new mandatory colourspace chunk precedence rules. That's a bigger issue because the current non-conformance in libpng stops apps implementing the new rule. libpng zaps chunks it shouldn't zap under the v3 rules and also introduces chunks in a way that may break app handling.

The good news is that it's what you like; a vast simplification and much removed code. The bad news is that's it's also a bigger change so .46 shouldn't be delayed.

[ctruta]

Changes: corrects extent of "int error" in png_XYZ_from_xy and, thereby, corrects the incorrect "declaration after statement" bug that I introduced. Also corrects the indentation and line-width errors that introduction of the new extent generates; stupid GIT.

Reviewed, approved, integrated, thank you.

Corrects seven spurious semicolons.

Reviewed, approved, integrated, thank you. Just so you know, this is a failure with all C language levels, not just C89. I added the option -pedantic-errors in a bunch of makefiles that we are already running on the CI.

For not-so-good reasons PNG_FIXED_EXPORT and FP_FP_EXPORT provide their own semicolons. I forgot. My bad. Fixed.

Those reasons no longer apply to libpng18, though... Hint... Hint...

Introduces: spurious end-of-line spaces in aclocal.m4. I blame autoreconf, well, I blame git too - git commit -s -a. I don't much care about this latter.

You are correct not to care, but the tooling that I use for libpng maintenance does care. (For example, linters which I haven't integrated into ci/ci_lint.sh, as well as diffing tools, etc.) I used to strip the trailing whitespaces -- I mean, my tools used to -- but now I pacified them. When I have some spare cycles, I might submit a nit-picking patch to GNU autoreconf.

[ctruta]

I also have more PNGv3 conformance stuff connected to the breakage caused by the new mandatory colourspace chunk precedence rules. That's a bigger issue because the current non-conformance in libpng stops apps implementing the new rule. libpng zaps chunks it shouldn't zap under the v3 rules and also introduces chunks in a way that may break app handling.

Acknowledged.

The good news is that it's what you like; a vast simplification and much removed code. The bad news is that's it's also a bigger change so .46 shouldn't be delayed.

Also acknowledged, with the agreement that the good news are indeed good and the disagreement that the bad news are not bad at all.

[jbowler]

OOOH neat; the commit ID is almost "deadsea". Neat, let's work on this; commit ids to attain:

deadsea (of course)
deadant
deasire (hey, that's a good one.)

[ctruta]

I haven't had any 0xbadf00d yet.