It's been almost two years since I started doing Windows Touhou patching, it's about time I actu...

Yeah, I suck at inventing new terminology. Just read the comment, it should make more sense then.

Yeah, it feels sort of weird to export these as part of win32_utf8, but hey, thcrap needs that f...

... or none at all. For now, this is mainly to get rid of the unnecessary TextOut() calls when c...

Again, thanks to Kill-o-matic for convincing me to ultimately get this done. I really managed to...

Yup, we're finally about to address that old TODO.

Any tab alignment in tokens that only use font commands is confusing and unintuitive, so let's g...

Introducing, at last, a state structure! Well, we had to take this step one day...

Seems like we've never had a test case for this, until we've encountered th07 and std6txt.anm.
T...

With all the progress we've made and th143 introducing a new font, that font replacement comment...

It's just another normal day of development. Today, a new set of detours are on the agenda. Ever...

We do need those to ensure locale independence in DLL injection chaining after all. As a result,...

Turns out that a certain modder needs to wipe everything below his patch.
Really though, this sh...

Yeah, haven't had one of *those* in a while.

After changing the variable names to better reflect their actual usage and removing unnecessary ...

... yeah, why?

Even less duplication! \o/

5 parameters, one of them wrong... OK, too much duplication.

Come on, this thing is global anyway.

Also renaming the variables since it hasn't been obvious enough.

Turns out that we might like to completely overwrite objects with other data types.

Eh...?!? Apparently, the "DWORD-aligned" part in MSDN's documentation of ReadDirectoryChangesW()...

Ah, the old, ancient pearls you find by just randomly reading in your code.

Yes, I totally did not have this idea earlier. :/

... and just when you thought you had a rock-solid build, the one main thing you wanted to focus...

Apparently, raw.githubusercontent.com automatically does this. As a result, the checksums create...

I spare you the long, drawn-out history of how this came to be. After months of research, fighti...

Who would have thought that basic string templating would be that trivial?

Apart from needing i...

Seriously, *this* had to become the industry standard container format for Deflate? Why.

... because json_object_set_new() dereferences the new value if the containing object doesn't ex...

Yeah, it also took me a while to wrap my head around the concept of returning zero (= 1 value) f...

The first step to a saner init system. This guarantees that every core module (especially string...

With detour chaining, this practice is not exactly conducive to good style.

Gosh, this was *so* aggravating.

I've come to like this format quite a bit.

Whoops, forgot this part during the refactoring. This actually caused crashes when dumping the s...

Fixes a bug introduced with advanced image patching in 2014-01-03. ANM entries can define sprite...

One more repatchable file, and one less instance of hardcoded initialization in thcrap_init().

Previously, this effectively did not render any token beyond the first one that didn't provide a...

In conjunction with the jsondata module, this implements hot-repatching (issue #5) for global data.

Makes the code clearer, is marginally faster, and gets rid of some warnings.

This also temporarily removes spell repatching.

This allows us to run a full detour chain outside of its own execution, which will be necessary ...

Well, turns out that png_image_write_to_file() internally uses fopen(), which in turn internally...

No point having this in the API when JSON strings are almost always a better alternative to a pl...

That has been a generic image resizer all along.

We'll be needing that functionality over at the PC-98 side.

Also, who needs return values anyway.

Well, we had to enter CRT territory at one point.

Very important, detour_next() would otherwise loop infinitely.