This is a somewhat intricate problem. As long as we linked with --no-as-needed,
these problems could not manifest themselves, since all dependencies are spotted
correctly by SCons and thus added as direct children of the executable.
But when we switch to --as-needed linking, the linker will omit some of
the dependencies given from the build system, when the code to be linked
doesn't call directly into these dependencies. But of course dynamic modules
may depend on each other, and indeed, the Lumiera libs do so. Thus
the linker may omit the dependency to liblumierasupport, and just add
a dependency to, say liblumierabackend. But the backend in turn
depends on the support library.
Now the problem is, that when resolving several steps deep into such
a dependency chain, our special relative path resolution scheme fails.
The fix is to give each lumiera module itself another relative path
resolution spec, which overrides at that point the root spec given
for the executable. Thus, we define
- for the executable: "search at $ORIGIN/modules"
- for the modules: "search at $ORIGIN/../modules"
This accounts for the fact, that a module, which is the Origin
for a transitive resolution step, already sits in a subdirectory
below the executable; thus step one level up and devle down into
the hard wired modules directory. Alternatively, we could also
use just "search at $ORIGIN" (i.e. in the same directory).
But assuming that in future we'll roll several core plugins,
which also count as "Lumiera modules", the scheme defined here
is more flexible, since it allows to place those core plugins
into sibling directories.
we still don't check for the precise boost version number in the
SCons configuration. We might do so in future, but for now our
policy is that precise version dependency checking is the business
of the packager (i.e. the debian package). The SCons build just
has to ensure the absolute necessary baseline
Why GNU extensions? They where on by default previously,
so we're changing nothing besides the C++ standard level.
AFAIK, we're using a GNU extension at one place, and this
could be replaced by 'decltype' now.
This improves sane library dependencies, as mandated
by debian policy. Some of the librt functions are in such
wide use, that we'd get them through transitive dependencies
anyway. But linking explicitly against a function adds the
DT_NEEDED for this library to our build artefacts. This is
correct, since we're actually using these functions ourselves,
making them into direct dependencies.
The recommendation is to use the link flag --no-undefined
and to fed *all* dependencies to the respective link step.
This changeset enables this strict linking of dependencies.
It turned out that our dependencies were already sane
(with the sole exception of a direct dependency to X-Lib
in the XV viewer widget)
The XV-Viewer widget in our GUI uses four direct calls
to the X-Lib. This was discovered by strict dependency checking,
as mandated by new Debian policy
...this will be the second preview release
Lumiera is still in pre-alpha stage, and thus there
are no proper releases, just preview snapshots
from time to time.
But we're providing Debian packages allready
- upgrade the configuration to a current version
- provide a frontpage with cross-links to other documentation
- define a set of modules; relevant classes and files can be
added to these, to create a exploration path for new readers
- fix a lot of errors in documentation comments
- use a custom configuration for the documentation pages
- tweak the navigation, the sections and further arrangements
Note: this drops some backwards compatibility. We're targeting now
roughly the range between Ubuntu-Precise (LTS) and Debian/testing,
with Debian/stable as the reference system.
The naming scheme for Boost-Libraries was adjusted with Boost-1.42
for Unix-Platforms. Now the '-mt' suffix isn't included any more, but
the libraries available through the usual packaging mechanisms can be
assumed to be thread safe.
See also http://issues.lumiera.org/ticket/759
tests used to be defined ad hoc and test definitions
are scattered confusingly over various directories.
Now built some simple rules into the buildsystem
to allow organising the tests into layers and
linking them accordingly.
Note: this switches to building shared objects
for the test classes too, which effectively speeds up
both re-building and re-running of test cases
* rfcs are now all stored in the rfc folder (formerly hold the final rfcs)
* a new rfc_final folder is created
* the state folders (rfc_final|pending|parked|dropped) now contain
symlinks back to the rfc pool
the icon rendering script from Joel had a function
to invoke Inkscape; it seems this was a leftover
from earlier attempts to render the icons.
Now, we seem to rely on lib Cairor solely
Not sure if this works; problem is that the icons
generated from SVG are just dumped into the target
folder, but we miss to generate the respective
installation targets
- use custom builders
- clean up specification of target paths
- generated executable is fully relocatable
- read a bootstrap INI instead of compiled in searchpath
