Add the ability to write to the compressed stream directly from
files::images.
Add streaming operator support which is always in a standard byte
format in the output image.
Remove archive_begin/archive_end calls when loading symbols. The symbols
reference object files and they may reference archive and so the archives
need to be open and available. The archive begin/end must be handled at
a higher level in the program.
Fixed image initialisation where some constructors did not initialise
all class variables.
Add section support to the rld::files namespace. This allows section
information to be added to an object file without the ELF file
needing to available therefore removing the need for object file
remaining open all the time. These sections are a copy of the
information and can be copied and moved as needed. This is used
to make layouts.
Added support for an RTEMS RAP format application file. The format is:
<header>
<LZ77>
<Application Script>
<[1..n] ELF Object files>
</LZ77>
Where the header is a text string of fields delimited by ',' and terminated
with a line feed (\n). It is variable length:
RTEMS-APP,0000000,01.00.00,LZ77,00000000\n\0
where:
RTEMS-APP : file tag for quick acceptance and rejection
Length : the length of the application in bytes including the
: header
Version : Version of the application format.
Compress : The compression format.
Checksum : CCITT CRC32 checksum.
Following the header is a nul ('\0') character then an LZ77 container
with the application loader script followed by the ELF object files.
Note, the script format will be documented else where.
Note, the final version may add a 32bit length field before each part
in the compressed container to delimit the size of the file to be
read. This is currently not in this version.
Fix the finding of a file name in the GNU extension for long names
in GNU archives so the correct location is referenced.
Made the image read and write routines keep reading if not all the
requested data is read or written due to possible signals.
This also adds support to the ELF classes that wrap libelf. While
this is now done and seems to work I will not be using an ELF
file to hold the metadata after all.
The refactoring allows better reuse of the ELF support and cleans up
some hacks from the generic file and archive handling improving the
separation of the file handling from the file format, ie ELF. The
handling of ELF object files and ELF object files inside archives
is cleaner.
The refactor cleaned up the symbol handling where the symbols now
reside in the ELF file object and references are take in symbol
pointer containers and symbol table containers.
The main purpose of the refactor is to allow support for creating
and writing ELF files.
Also added an rtems-syms command where special symbol support
can be added.
If an object on the command line depends on another object the output
code wrote the object files and then the dependent files and a
command line object file that is dependent ended up in the object
and dependent lists. A simple merge and unique fixed it.
Also moved the script generation code into a separate function
that can be used in a application container.
Add a -C (also --cc) option to allow the CC to be used when linking to be
provided by the user rather than using the path. This support allows user
who work with the full path to tools rather than the environment to make
use of the linker without them needing to play with environment table.
Rename rld-gcc.[h.cpp] to rld-cc.[h,cpp] because gcc may not be the
only compiler/linker used by the RTEMS project.
* specbuilder/specbuilder/build.py,
specbuilder/specbuilder/crossgcc.py,
specbuilder/specbuilder/defaults.py,
specbuilder/specbuilder/linux.py,
specbuilder/specbuilder/spec.py: Add CentOS support for older
Pythons. Add options to build the tools with specific flags.
