libjpeg uses unsigned ints for data length, so we must use size_t
everywhere.

in UCS table building

WHen loading a vips image with Coding set, make sure that Bands and
BandFmt are correct.

Now sets 0 rather than inf.

sharpen with sigma 0.5 was doing nothing, since the new int precision
rules meant that we genera...

Expand note on @background for pdfload.

See https://github.com/libvips/libvips/issues/1412

We could pre-shrink so much that an axis went to 0.

See https://github.com/lovell/sharp/issues/...

giflib does not check this either :(

It seems giflib does no checking of image dimensions at all, not even
height == 0.

Earlier refactoring broke GIF map fetch.

Some old rad code could cause a read beyond end of buffer.

possible race in free

so pages and delay_length can't get out of sync, even with corrupt GIFs

lots of GIFs are very badly structured :(

Add the new eof handling to the memory reader.

Relaxing the read error rules made looping possible .. make sure we
always stop explicitly on eof.

They could be out of bounds enums. This used not to matter, but we use
them more now, so out of ...

Some GIFs don't follow the standard very closely. Let them through.

See https://github.com/libv...

some sample images for the file loader

oss-fuzz normally fuzzes with memory buffers, but a lot of the libvips
loadres, especially CSV a...

libheic versions change rounding behaviour

libvips could harmlessly read beyond the end of a string with a crafted jpg
file

now that relational const is fixed up

there's a problem with out of bounds values, for example:

vips relational_const k2.jpg x.v equ...

WebP loader: verify upper limit on dimensions in header

move fuzzing corpus into a single dir

add fuzzers for vips_smartcrop and vip_mosaic

argh rounding with different lib versions

now a switch/case pair

A neater solution! In Python it should be something like:

merge = pyvips.Image.switch([tests])....

gcc has an "unknown pragma" warning which triggers for clang pragmas we need to hide clang compi...

only 1/2 done

Tail processing in shrinkv had an implicit assumption of round-down, but of
course we round to n...

This reverts commit 263f69d5818cfe6235774577813f8de709e40b33.

This reverts commit 4dbf2167d0387ab8dfc11be6de227aef9c7c33de.

This reverts commit faafd4eee0d1cf5287992dbdbc57d29e0d9a6fba.

This reverts commit 88b3aa8f2b0b9af2f692c0be51adcb7f2aa348c5.

libvips has to use ReadImage() when it opens an image via MagickCore,
and this can allocate very...

things like 65494x5 pixel PNGs were being allowed through and caused
timeouts

fuzzed tiffs can give huge and mysterious values for tile_size

nope, fails on travis

getting strange parse errors on asan

since we now have four of them

they need to be in separate files

for lsan testing

We used to lock within documents, ie. we did not allow two threads to
work on the same file.

Ho...

See https://github.com/libvips/libvips/issues/1392