Files
Sepp J Morris 31738079c4 Upload
Digital Research
2020-11-06 18:50:37 +01:00

141 lines
4.9 KiB
Groff

.th ld8k 1 Loader/linker
.sh DESCRIPTION
ld8k - loader/linker
.sh SYNOPSIS
ld8k [option] file ...
.sh DESCRIPTION
Ld8k combines several object programs into one. It is capable of linking
both segmented and non-segmented Z8000 object modules. However, all of
the modules linked must be of the same type. Ld8k also has facilities for
searching libraries in the format prepared by ar8k.
Normally, the output (default = x.out) is prepared for execution, however
the -r option may be used to preserve relocation information so that the
output can be used as further input to ld8k.
If no errors occur during the execution of ld8k, the output module is
marked with a magic word that indicates that it is executable.
Arguments to ld8k are processed in the order that they appear in the
command line. Libraries are searched exactly once at the point at which
they appear in the command. Only routines that resolve at least one
external reference are loaded. Therefore ordering of libraries is
important.
Whether segmented or non-segmented linking is done depends on the
type of the first object module encountered, since no mixing of types
is permissible. In searching archives, only entry points in modules
having the correct type are considered. Hence archives can have modules
with duplicated entry point names so long as they are in different object
module types.
There are several options. Except for library specification (-lxx),
all should appear before any file names.
.in+5
.ti-5
-i\ \ \ Prepare the object file for "split I and D" space. That is,
start the data addresses at zero, and in the case of segmented loads,
segment numbers can also be reused.
.ti-5
-lxx\ Search the library name "libxx.a".
This may appear anywhere in the command line.
.ti-5
-n\ \ \ Mark the text read only so that it may be shared. This is
meaningful only for non-segmented loads. The data boundary is moved up
to the next 8K byte boundary.
.ti-5
-o\ \ \ The next argument is taken as the name of the object file
instead of "x.out".
.ti-5
-r\ \ \ Preserve relocation information in the output even if all references
are resolved.
.ti-5
-s\ \ \ "Strip" the output by removing the symbol table and relocation
information to save space in the object file.
.ti-5
-t\ \ \ The next argument is a decimal number that sets the size of the
stack segment.
.ti-5
-u\ \ \ Enter the next argument as an undefined reference so that loading
can be accomplished solely from an archive file.
.ti-5
-d\ \ \ The next argument is taken to be a file name containing segment
numbers and name correspondences. In this way specific segment numbers
can be assigned to named segments, or the loader can be forced to ignore
specific segment numbers. The are two formats for the descriptor file.
The first format is a sequence of ASCII lines as follows:
<number>[,<typechar>]*=<name>[,<name>]*
or
<number>
The first form requires the named segments to be assigned the given
number. The <typechar> should be one of the letters "t", "b", "d", or "c"
to indicate segment types text, data, bss, or constant respectively. If
the type is not indicated, the type of the named segment will be used.
If all of the named segments with the same name
will not fit into 64KB, an error
message will be issued. The order in which the named segments are assigned
to the numbered segment is the order in which the names appear in the
control line. The second form effectively reserves the segment number and
does not permit the loader to use it.
The second form of the descriptor file is as follows:
<type>=<number>[-<number][,<number[-<number]]*
In this form the type is one of the names: "text", "bss", "data", or "cons",
and the numbers are the valid numbers for this type of segment.
In both forms, the output file will have the segments appearing in the
same order in which they are first encountered in the descriptor file. In
this way, the object order can be forced. For example:
.nf
.in+10
text=11,9,7
cons=10,12,8
data=15-18
bss=3-1
.in-10
.fi
will cause text to be put first in segment 11, then 9, then 7 (as each of
the segments is filled in turn), constants into 10, 12, and 8, data into
15, 16, 17, and 18, and bss to be filled into 3, 2, and 1. The specific
ordering will be preserved in each case. If this proves to be an insufficient
number of preallocated segments, segments 0, 4, 5, 6, 13, 14, 19+ will be
selected in turn. (Remember that segments can be avoided using the first
form above.)
If segment types are mixed (text=5/data=5 for example), both types of
segments will be allocated first in segment 5, then subsequent allocation
will be separated. However, if the file reads text=4,6/data=5,6, both
data and text will be allowed to spill into segment 6.
.ti-5
-d"descriptor line" this is a shorthand form of the preceding control
in which one line will suffice. -d commands may be repeated, in which
case they are processed in the order they are encountered.
.in-5
.sh FILES
lib*.a libraries
x.out object file