Intentionally left blank

Here you can see the more detailed dependency information for the target `src/ban.o` that was obtained using `strace`. The target will now also be rebuilt if the compiler changes or...

The argument to `redo-dot` is a shell pattern for nodes to include in the graph. Every other node in the full dependency graph that is connected to ta node matching...

It seems to me that using a precompled header the way I did it introduces a lot of bogus dependencies. An example: * A file `foo.cpp` depends on `foo.h`. *...

I have just confirmed that using precompiled headers make dependency checks for incremental builds *much* slower. I have also confirmed that parallel builds work.

Setting the hash function used for dependency checks (that happen if the inode timestamp does not match what is on record) to `wc -c` instead of `md5sum` reduces the time...

To use precompiled headers you now have to do `redo src/precompile.h.gch` before building. If the precompiled header exists at build time, it will be used and recorded as a dependency....

When you change build scripts, you might want to verify the dependency graph. For the last commit, I used `redo-dot` to output the build graph with and without the changes,...

Parallel builds seem to work fine. On a Thinkpad X390: * A full rebuild using `redo -j 8` takes 7 minutes and 50 seconds. * Invoking `redo -j 8` afterwards...

By fixing a small bug in a build rule I could make [Avery Pennaruns redo implementation](https://github.com/apenwarr/redo) build Minetest using these dofiles. I found that on the Thinkpad X390, apenwarr-redo takes...

> > * It seems that apenwarr-redo is timestamp-based. > > That's mentioned here, isn't it? http://news.dieweltistgarnichtso.net/bin/redo-sh.html#implementation-comparison Indeed – I simply forgot about it. I guess that then fully explains...