On 2026-06-11 19:10 +0200, Sven Joachim wrote:
> Source: unibilium
> Version: 2.1.1-2
> Severity: important
> Tags: forky sid
>
> Yesterday I uploaded a new ncurses patchlevel to experimental, and the
> unibilium autopkgtest fails with ncurses-base 6.6+20260608-1, see [1]:
>
> ,----
> | 41s # Parsing /usr/share/terminfo/x/xterm ...
> | 41s not ok - db entry for xterm is too small to be valid
> | 41s # Parsing /usr/share/terminfo/x/xterm-256color ...
> | 41s not ok - db entry for xterm-256color is too small to be valid
> `----
>
> I have not looked at the details yet, but I suspect that the problem is
> just the opposite of what the error message claims, namely that the
> entries are too _large_ to be considered valid. The two terminfo
> entries exceed the size of 4096 bytes, and libunibilium apparently has a
> problem with that.
Looking at the source now, this is indeed the problem. The test calls
unibi_from_file() on the terminfo entries, and that function calls
unibi_from_fd() on the file descriptor for the file which
unibi_from_file() just opened.
As mentioned in the documentation (see doc/unibi_from_fd.pod),
unibi_from_fd(fd) reads up to 4096 bytes from fd, and that is not enough
if the terminfo entry is larger.
> Note that with the new format introduced in ncurses
> 6.1, compiled terminfo entries may be up to 32768 bytes.
I suspect the solution is to enlarge MAX_BUF from 4096 to 32768 in
uniutil.c (and adjust the documentation in doc/unibi_from_fd.pod), I
leave it up to the maintainer to test that.
I also did a quick test on neovim with TERM=xterm-256color. The good
news is that it starts up and seems to work, but there might be subtle
bugs because unibilium did not read in the complete file.
Cheers,
Sven
