aboutsummaryrefslogtreecommitdiff
path: root/documentation/PROBLEMS
diff options
context:
space:
mode:
Diffstat (limited to 'documentation/PROBLEMS')
-rw-r--r--documentation/PROBLEMS70
1 files changed, 70 insertions, 0 deletions
diff --git a/documentation/PROBLEMS b/documentation/PROBLEMS
new file mode 100644
index 0000000..a07fbda
--- /dev/null
+++ b/documentation/PROBLEMS
@@ -0,0 +1,70 @@
+STRUCTURE
+
+
+VARIABLES
+
+%% It would be nice to be able to include makefile in a way that
+ variables are restored.
+
+@% I definitely need some naming convention for "global" variables.
+ Since I include potentially unknown makefiles (like library
+ dependecies) I cannot predict what they do. This is another
+ argument for variable framing.
+
+ Some ideas:
+
+ - Variables that start and end with '%' are considered global
+ and system. Subsystems may define such vars if there is
+ a real need. (Currently defined names should be documented
+ here. Each subsystem should document names that it defines.)
+ Subsystem should prefix each such var with its name (e.g. cxx_).
+
+ - Variables that contain '-' are considered 'functions' and not
+ framed. Generally subsystems should define those. The names
+ should be chosen carefully in order not to cause name clashes.
+ (Each subsystem should document names that it defines).
+ Subsystem should prefix each such function with its name
+ (e.g. cxx-).
+
+ - Variables that appear in %frame_exclude% are not framed. General
+ rule is to add only special one-word functions (usually defined
+ by base subsystem) that are used often and thus having one-word
+ terse name is very important; for example 'include' or 'import'
+
+
+%% After the inclusion of dependency makefiles I cannot use src_base
+ becasue it was reset by the included file. So I am forced to use
+ src_root now which will hurt especially sore when I include from
+ sub-directories.
+
+
+
+DEPENDENCIES : STRUCTURE
+
+@@ It would be nice to be able to chose dependency type project-wise
+ and on a by-makefile basis. This could be tricky, however.
+
+@@ Missing headers assumed generated with some inflexible logic (cur dir).
+ This can also vary among compilers (which is not something I would worry
+ too much as long as g++ works). Can I make it more reliable? Maybe capture
+ this (or essential) dependecies in a Makefile?
+
+@@ Can I specify dependencies in CORBA projects manually. Do I want to do
+ that?
+
+@@ Don't include dependecies when cleaning (configuring?).
+
+
+
+MISC
+
+@@ If I have generated idl files in src directory then build in some other
+ directory will fail or will use wrong headers.
+
+%% .NOT_DEFAULT: target in make?
+
+%@ configuration targets should not be parallel. try make -j 4 Maybe I can
+ detect that those files do not exist and disable parallelism for this
+ run. See .MUTEX bug (#3656).
+
+@@ relative paths and pattern matching does not work very well together