I've been working on adding a new linker mapfile check
(mapfilechk) to cadmium, patterned on cddlchk, as detailed
in the CR:
6785284 Mapfile versioning rules need to be more visible to gatelings
In a nutshell, there will be a standard comment in all of the OSnet
link-editor mapfiles, and the mapfilechk command will be used to
keep them intact. I already have it working, largely by copying
cddlchk, and modifying it to suit. However, I've hit a problem
with cdm.py that I'd appreciate some help with.
mapfilechk examines any file with a name that matches '*mapfile*',
and ignores all others. However, there are a small number of files
in OSnet that match this pattern that are not actually mapfiles.
For instance, mapfilechk itself. I need to have an exceptions list
for these files. I am using the .NOT file mechanism supported in
cdm.py via the not_check() function. Using cddlchk as an example:
def cdm_cddlchk(ui, repo, *args, **opts):
...
filelist = opts.get('filelist') or _buildfilelist(repo, args)
...
exclude = not_check(repo, 'cddlchk')
for f, e in filelist.iteritems():
if e and e.is_removed():
continue
elif (e or opts.get('honour_nots')) and exclude(f):
ui.status('Skipping %s...\n' % f)
continue
...
The problem I'm encountering is that the list of files returned by
_buildfilelist() is relative to the working directory, while the
list of file path exceptions needs to be relative to the workspace
root. Consequently, exclude() only matches if my working directory
is set to the workspace root.
In my exceptions (.NOT) file, I currently have:
usr/src/lib/README.mapfiles
usr/src/tools/scripts/mapfilechk.1
When I run my mapfilechk command while my working directory is
usr/src/tool, I get:
% hg mapfilechk
Mapfile comment check:
Warning: No mapfile header block in file scripts/mapfilechk.1
Warning: No mapfile header block in file ../lib/README.mapfiles
which is wrong. If I move to my workspace root however:
% cd ../../..
% hg mapfilechk
Mapfile comment check:
Skipping usr/src/lib/README.mapfiles...
Skipping usr/src/tools/scripts/mapfilechk.1...
which is correct.
Is the use of working directory relative paths for the
exclude mechanism intentional? It prevents the use
of path components in the .NOT files, which is a pretty
big limitation.
For the moment, I'm just testing from the root of my workspace,
but I can't integrate it this way. I'd appreciate some help with
finding a solution that is not working directory dependent.
Thanks...
- Ali
