Now binaries can be (cross-)built using GCC for 32- and 64-bit Windows.
Additional improvements:
- Within MSYS2, try to install the required dependencies automatically.
- Within MSYS2, fix passing libdir paths to pkg-config.
- Prune documentation from extracted package files,
addressing the incredible slowness of Windows filesystem operations.
- Fix the script name in README.adoc instructions.
This makes use of our image processing capabilities in order to
turn arbitrary image files into normalized thumbnails,
upload them to a temporary host, and pass the resulting URI
to a search provider.
In future, fiv should ideally run the upload itself,
so that its status and any errors are obvious to the user,
as well as to get rid of the script's dependency on jq.
wperl is necessary to get rid of the console window,
which is merely one of several issues with the PAR Packer-based
ExifTool bundle used in the last commit.
The Perl installation could be heavily trimmed down,
but it seems to require a very manual process.
- Fix launching of subprocesses (missing gspawn helpers).
- Discard unused GSettings schemas.
- Make the program find its user guide.
- Bundle a somewhat suboptimal version of ExifTool.
There is no point in claiming speed, it turns out to be a strange focus
to have, considering the amount of available innovations to make.
The new name does not appear to be taken by anything important.
The gdk-pixbuf plugin does not work here, for whatever reason.
Moreover, close integration exposes higher bit depths, metadata,
and auxiliary images.
The library is awful and copylefted, but it's the only reasonable
thing that works.
Sadly, they don't have a canonical extension, and they don't show up
in the browser. We might want to employ some level of sniffing.
The first 16 bytes are enough to identify a lot.
Speed matters here, and this makes us about 20 percent faster
at loading large directories.
Moreover, libpng's PNG_ALPHA_BROKEN is indeed broken.
Thumbnails have a fairly fixed format, so there are very few practical
corner cases that could have been missed.
We need to refactor, so that SVGs are pre-rendered on each change
of scaling by librsvg directly, because some elements may be rasterized.
It would be best to also support building against resvg.
