| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
To force a rebuild of the docs, run
$ make -C manuals_and_examples clean all
|
| |
|
|
|
|
|
|
|
|
| |
Renames a few programs which are called the same as a directory.
We need to change the directories later, and rename them back.
Lacks qlparser and rasdl stuff, as those use flex and bison and
thus need special handling.
|
|
|
|
| |
These prevented Automake from creating the correct build rules.
|
| |
|
| |
|
| |
|
|
|
|
|
| |
We still need to adapt the actual make rules and vars
so that an actual build might work.
|
|
|
|
|
|
|
|
| |
We will keep only three Makefile.am:
* $(top_srcdir)
* manuals_and_examples/ (needs code to be generated first)
* systemtest/ (needs everything to be built)
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Clean up the building of the doxygen docs for VPATH
builds and "make distcheck". This shifts the doxygen
run from "make install" time to "make all" time.
Fix a number of permissions in the docs we "cp -R" to
install them, and create the target dir before copying
manual files.
Have doxygen ignore subdirs like .git.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Reorganize the building of the rpcif* files such that we do
not ship these files in the dist tarball.
These generated files should not be in the dist tarball, as
* they can be regenerated easily
* they disturb our wildcard emulation based on the dist-hook
generating a list of files in a separate *.mk file.
At a later time, the rpcif stuff should be moved to a place
where both clientcomm/ and servercomm/ can equally trigger a
rebuild.
|
| |
|
|
|
|
|
|
| |
Do not generate java/Makefile and petascope/Makefile.
Those appear to be unused and thus not shipped in the dist tarball.
|
|
|
|
|
|
| |
This requires newer automake versions, and xz being
installed. However, with those HUGE dist tarballs of
5+MB gzipped, this makes a lot of sense.
|
|
|
|
|
|
|
| |
Support (srcdir != builddir) VDIR builds.
This will also help a lot with preparing for a working
"make distcheck".
|
|
|
|
|
| |
Also make sure that the list of files is updated on
"make dist".
|
|
|
|
|
| |
Move automake option "--foreign" from Makefile.am to
configure.ac.
|
|
|
|
|
|
|
|
|
| |
automake 1.10 was released around the time autoconf 2.61
(which we already require) was released in late 2006, so
that should not present a real obstacle to people.
However, it at least avoids compatibility issues with even
more ancient versions.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The CPPFLAGS, CFLAGS, CXXFLAGS, LDFLAGS vars should be
free for use by the person who does the build, so rasdaman
itself should put all its internally needed flags to
AM_CPPFLAGS, AM_CFLAGS, AM_CXXFLAGS, *_LIBS instead.
We do just that, and also put all the C preprocessor
defines in _CPPFLAGS so they are used for compiling
both C and C++ source files. This redistributes some
responsibilities from the _CFLAGS and _CXXFLAGS vars.
Also use _LDADD _LIBS where appropriate.
As a result, the user can now properly set CPPFLAGS,
CFLAGS, CXXFLAGS, LDFLAGS at configure or make time
and have these values used in the build.
|
|
|
|
| |
Also make sure the "cd" succeeds before trying to run make.
|
| |
|
|
|
|
|
|
| |
Those shell globs are only used in shell globs anyway, so defining the
make var to contain the actual file names actually makes no difference,
and having the make var contain only the globs works just as well.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Redefine $(pkgdatadir), $(pkgincludedir) and $(pkglibdir)
in order to make the subdir component configurable with the
RDM_SUBDIR variable.
Then install all libraries and include files to the package
specific subdirectories, as their names are far too generic
to allow general system wide installation, such as e.g.
"bool.h" or "libnetwork.a".
Move some very custom and non-general scripts (some with generic
names) to a new $(pkglibdir)/bin/ subdir.
|
|
|
|
|
|
|
| |
Place $(logdir) in $(localstatedir)/log/$(PACKAGE) by default.
$(prefix)/log is not used by anybody, and special-casing a prefix of
"/usr/local" for a logdir of "/var/log" without "local" makes no sense.
|
|
|
|
|
|
| |
The rasmgr.conf file is already installed into (and used in)
$(sysconfdir) aka -DCONFDIR="...", so there is no need to install it to
$(bindir).
|
| |
|
|
|
|
|
| |
This silences a g++ warning, probably without endangering an actual
code change.
|
|
|
|
|
| |
Update INSTALL file with instructions on initializing the build
system, i.e. on how to run autoreconf.
|
|
|
|
|
| |
Showing build system generated files in the "Project Explorer"
makes no sense.
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Fix typo for test port definition: The defined variable
is called RASMGR_PORT, not RASMGRPORT.
|
|
|
|
| |
CHECKLIMITS is never defined, and never used, so remove it.
|
|
|
|
|
|
| |
RMANBASE is defined in configure.ac to be ${abs_top_srcdir}
already, and BASEDB(CXX|LD)FLAGS and RASMGR_PORT are also
AC_SUBSTed.
|
| |
|
| |
|
|
|
|
|
|
|
| |
The #ifdef preprocessor directive only accepts a macro name,
not a complete expression.
For expressions, use "#if defined(foo) || defined(bar)".
|
|
|
|
|
|
|
| |
The function's return value type should be QtONCStream* instead
of QtONCStream::QtONCStream*.
This makes g++ actually compile qtdelete.cc.
|
|
|
|
|
|
| |
The C preprocessor cannot properly handle the dot in the
"#define _RPCIF.H_H_RPCGEN" macro rpcgen creates, so we change
that to "#define _RPCIF_H_H_RPCGEN" instead.
|
|
|
|
| |
"dnl" is the safer bet for comments in configure.ac.
|
|
|
|
|
| |
Defining "clean" target in an Automake build system is wrong.
Use "clean-local" instead.
|
| |
|
|
|
|
| |
Properly quote m4 args, and avoid confusing it with bison.
|
|
|
|
| |
Use proper quoting for shell variables with spaces in their values.
|
|
|
|
| |
Document the "git clean" trick (for now, more to follow).
|
| |
|