fossil: fix #29591, don't build windows specific files, remove -O2 CFLAG

[ci42]

changes:

• include upstream fix for #29591
• don't create the windows specific make files
• remove -O2 from ${configure.cflags} as it is inconsistence with the -Os flag set by the makefile

[ryandesign (Ryan Carsten Schmidt)]

Replying to ciserlohn@…:

• remove -O2 from ${configure.cflags} as it is inconsistence with the -Os flag set by the makefile

I see that they use -Os, and then MacPorts adds its -O2 after, which should mean -O2 takes precedence. The port seemed to build fine with -O2... are you saying the program will work better with -Os?

If so, I think the preferred way to indicate this is not:

configure.cflags-delete -O2

but instead:

configure.optflags  -Os

[ci42]

do not touch optimization flags

[ci42]

set ${configure.optflags} to -Os

[ci42]

delete -O2 from ${configure.optflags}

[ci42]

Replying to ryandesign@…:

Replying to ciserlohn@…:

• remove -O2 from ${configure.cflags} as it is inconsistence with the -Os flag set by the makefile

I see that they use -Os, and then MacPorts adds its -O2 after, which should mean -O2 takes precedence. The port seemed to build fine with -O2... are you saying the program will work better with -Os?

Not that I'm aware of. It is only a precautionary measure to not hit problems with -O2 as the upstream developer test their versions with -Os. Additionally they strive to avoid bloat. I think -Os is well-suited for that goal - the actual file sizes are 1.5M for -Os vs. 1.6M for -O2 and 2.8M vs. 3.0M for the universal variant.

If so, I think the preferred way to indicate this is not:

configure.cflags-delete -O2

but instead:

configure.optflags  -Os

Yes, configure.optflags is the right place, but wouldn't be

configure.optflags-delete -O2

the right thing to do. Otherwise we would add -Os twice.

Anyway, I added patches for all three options, feel free to pick the one you think is the most appropriate.

[ci42]

fixed in r79813