blob: fa25c6d5b52d626fe46e272f87db787b2bd38fe6 [file] [log] [blame]
xdc-D22 (corevers:16.0.2.0; B:200,G:200,R:200)
eXpanDed C Tools (XDCtools)
This tree contains the RTSC build and configuration engines for
target content. The tree is used in XDCtools 3.50 product releases. This
series of trees breaks compatibility with XDCtools 3.32 and older.
Librarian repo creation
-----------------------
The xdc.git repo was created to "track" the repo maintained at
eclipse.org and was created as follows:
git clone --bare ssh://git.eclipse.org/gitroot/rtsc/org.eclipse.rtsc.xdccore.git xdc.git
cd xdc.git
git config remote.origin.fetch "+refs/heads/*:refs/heads/*"
BUILD
o setup ssh to enable "git" from xlibrary as "dr" (if necessary)
cd ~xlibrary/.ssh
cp -f config-dr config
o fetch the latest changes from the Eclipse git server
cd xdc.git
git fetch
o Build this tree from Linux using standard SM-MAKE
# clone the tree and merge/review user content
ggentree -u sascha@ti.com xdc.git/
cd xdc-D??
# unzip build.zip from the last released tree
unzip ../xdc-D??/build.zip
# make sure the user email and name are your eclipse credentials
git merge origin/xdccore-D-dr
# build the tree using LSF to force older GLIBC to be used when
# linking Linux executables
SM-MAKE .clean
SM-MAKE .all-files >& Make.log
# update README
emacs README
git commit -a -m xdc-D??
# release the tree
greltree -m "something brief and meaningful"
o to make both the UNIX and PC sides, from a Linux workstation, go to the
top directory of the tree and type:
SM-MAKE .all-files
To clean the tree, type:
SM-MAKE .clean
o to build only packages in src/packages from scratch:
SM-MAKE .xdc-packages
o to build bundles:
SM-MAKE .bundles
o to run regressions:
SM-MAKE .regress
==============================================================
ADDITIONS
DELETIONS
CHANGES
xdc.rov added new functionality for reading objects pointed to by a pointer
and reading any memory location as it is of any chosen type
xdc.runtime Assert_isTrue and the macros invokes from it had several MISRA
issues, where integers were used in place of booleans, constants
were signed instead of unsigned, etc.
unsigned constants now use 'U' instead of 'u'
Registry used an internal signed int variable where an unsigned
short was enough, and the change fixed a couple of MISRA warnings
In Text.c, there is an operation with two unsigned short operands,
which are automatically promoted to ints or unsigned ints depending
on the sizes of shorts and ints. On Windows, unsigned shorts can be
represented with signed ints, so the promotion causes the operation
to produce a signed int as a result. This causes a warning because
that result is then compared to an unsigned int.
The same problem should have happened on Linux, but the compiler
did not emit the warning. This may need further research.
MRS