remove duplication in Makefile

[DerDakon]

Basically the same what we have already done for catman targets.

[schmonz]

Works for me with BSD and GNU makes. Nice incremental step in the direction of #52 (and perhaps in the direction of #28).

[josuah]

I thought the % makefiles rules were not portable because they are not POSIX, but they are at least working on BSD make : https://notqmail.z0.is/patch/0-patch/85555302dd1f6688babb236eb364e9e6f08c96ac/index.make.out

[josuah]

With these changes, along with multiple others, it looks like we are going more toward hand-maintained Makefile.

As the total number of files decrease (by using libc more), it might become more and more feasible to do so.

Is that the direction we are taking?

[schmonz]

My interpretation was the opposite. Interesting. I was thinking anything that highlights patterns in our existing dependencies is a step toward automation. But I’m not actually using redo, and you are. Does this change feel counterproductive to you, or not obviously worth the cost, or something like that?

[DerDakon]

Could someone please test if this actually honors both rules? I had the impression that BSD make was only looking at the compile-rule, but ignoring the later dependency rules.

[josuah]

Note that this removes dependency on headers: if they get modified and the project is recompiled, the associated .o might not be recompiled. Did nowaday NVMe disks made make clean all cheap enough?

This also introduces % rules that are not POSIX, but there are already common extensions used.

[josuah]

A slightly less readable but more portable way would be .c.o:, but I doubt this would support adding compile as dependency of .o.

[DerDakon]

Note that this removes dependency on headers: if they get modified and the project is recompiled, the associated .o might not be recompiled. Did nowaday NVMe disks made make clean all cheap enough?

This was basically my question. This is not intended, and it works with GNU make, but if BSD make doesn't support this I can scrap half of it because that is a regression.

[josuah]

https://notqmail.z0.is/commit/03b8497487b7ebc0580d92059ae77171fff71601/patch/0/log

Works without trouble on that instance.