fix(crop): prevent division by zero in margin calc

[tachibana-shin]

Change 1: In lib/koptcrop.c:92, the margin calculation computes:

margin = margin * (region->c2 - region->c1) / (kctx->dev_width - margin) / 2;

• When kctx->dev_width <= margin (e.g. very small device width or very large margin setting), the denominator becomes zero or negative, causing undefined behavior (floating-point division by zero).
• When the device width is too small to accommodate the requested margin, margin is set to zero (content fills the entire width).

Change 2: The file defines a global float min(float, float) at lib/koptcrop.c:32. This name collides with common min macros defined in system headers and other libraries, causing potential compilation issues or silent behavior changes depending on include order.

---

This change is 

[Frenzie]

Is there a reason we shouldn't we just use fminf() (from math.h) instead? The output seems to be identical at a quick glance.

Without the -ffast-math it just says branch to fminf which annoyingly doesn't tell me much about the internals. (I suppose I could check glibc etc.) According to the manpage NaN behavior may be different but in a way that sounds better.

These functions return the minimum of x and y.

If one argument is a NaN, the other argument is returned.

If both arguments are NaN, a NaN is returned.

[tachibana-shin]

Personally, I don't see any reason for it, but since this function is there, I don't want to delete it. The only difference between the results of the two functions is that k2_min returns 0.0 and fminf returns -0.0 when the boundary is -0.0 < 0.0, but I’ve looked through the code snippets and the parameters a + b are always greater than or equal to 0, so I don’t think there’s any difference.

[Frenzie]

I don't think that argument quite works if you're touching all the lines.

Pinging @hugleo, see #44.

Btw @benoit-pierre I noticed something funny if it bothers you or something:

libk2pdfopt/willuslib/math.c

Lines 26 to 27 in 1fcffeb

	#include <math.h>
	#include <math.h>

[hugleo]

I don't think that argument quite works if you're touching all the lines.

Pinging @hugleo, see #44.

It's fine with fminf.

[Frenzie]

Would you mind squashing the last two commits together?

[tachibana-shin]

Would you mind squashing the last two commits together?

yeah done