Compiled terminfo descriptions are placed under the directory \*d.
In order to avoid a linear search of a huge \s-1UNIX\s+1 system directory, a
two-level scheme is used: \*b/c/name
where
name
is the name of the terminal, and
c
is the first character of
name .
Thus,
act4
can be found in the file \*d/a/act4.
Synonyms for the same terminal are implemented by multiple
links to the same compiled file.
The format has been chosen so that it will be the same on all hardware.
An 8 or more bit byte is assumed, but no assumptions about byte ordering
or sign extension are made.
The compiled file is created with the
tic
program, and read by the routine
setupterm .
The file is divided into six parts:
the header,
terminal names,
boolean flags,
numbers,
strings,
and
string table.
The header section begins the file.
This section contains six short integers in the format
described below.
These integers are
(1) the magic number (octal 0432);
(2) the size, in bytes, of the names section;
(3) the number of bytes in the boolean section;
(4) the number of short integers in the numbers section;
(5) the number of offsets (short integers) in the strings section;
(6) the size, in bytes, of the string table.
Short integers are stored in two 8-bit bytes.
The first byte contains the least significant 8 bits of the value,
and the second byte contains the most significant 8 bits.
(Thus, the value represented is 256*second+first.)
The value -1 is represented by the two bytes 0377, 0377; other negative
values are illegal. This value generally
means that the corresponding capability is missing from this terminal.
Note that this format corresponds to the hardware of the \s-1VAX\s+1
and \s-1PDP\s+1-11 (that is, little-endian machines).
Machines where this does not correspond to the hardware must read the
integers as two bytes and compute the little-endian value.
The terminal names section comes next.
It contains the first line of the terminfo description,
listing the various names for the terminal,
separated by the `|' character.
The section is terminated with an \s-1ASCII NUL\s+1 character.
The boolean flags have one byte for each flag.
This byte is either 0 or 1 as the flag is present or absent.
The capabilities are in the same order as the file <term.h>.
Between the boolean section and the number section,
a null byte will be inserted, if necessary,
to ensure that the number section begins on an even byte (this is a
relic of the PDP-11's word-addressed architecture, originally
designed in to avoid IOT traps induced by addressing a word on an
odd byte boundary).
All short integers are aligned on a short word boundary.
The numbers section is similar to the flags section.
Each capability takes up two bytes,
and is stored as a little-endian short integer.
If the value represented is -1, the capability is taken to be missing.
The strings section is also similar.
Each capability is stored as a short integer, in the format above.
A value of -1 means the capability is missing.
Otherwise, the value is taken as an offset from the beginning
of the string table.
Special characters in ^X or \ec notation are stored in their
interpreted form, not the printing representation.
Padding information $<nn> and parameter information %x are
stored intact in uninterpreted form.
The final section is the string table.
It contains all the values of string capabilities referenced in
the string section.
Each string is null terminated.
Note that it is possible for
setupterm
to expect a different set of capabilities
than are actually present in the file.
Either the database may have been updated since
setupterm
has been recompiled
(resulting in extra unrecognized entries in the file)
or the program may have been recompiled more recently
than the database was updated
(resulting in missing entries).
The routine
setupterm
must be prepared for both possibilities -
this is why the numbers and sizes are included.
Also, new capabilities must always be added at the end of the lists
of boolean, number, and string capabilities.
Despite the consistent use of little-endian for numbers and the otherwise
self-describing format, it is not wise to count on portability of binary
terminfo entries between commercial UNIX versions. The problem is that there
are at least three versions of terminfo (under HP-UX, AIX, and OSF/1) which
diverged from System V terminfo after SVr1, and have added extension
capabilities to the string table that (in the binary format) collide with
System V and XSI Curses extensions. See terminfo(\*n) for detailed
discussion of terminfo source compatibility issues.
As an example, here is a hex dump of the description for the Lear-Siegler
ADM-3, a popular though rather stupid early terminal:
Some limitations: total compiled entries cannot exceed 4096 bytes.
The name field cannot exceed 128 bytes.