aboutsummaryrefslogtreecommitdiff
path: root/documentation/MAKE
diff options
context:
space:
mode:
Diffstat (limited to 'documentation/MAKE')
-rw-r--r--documentation/MAKE31
1 files changed, 31 insertions, 0 deletions
diff --git a/documentation/MAKE b/documentation/MAKE
new file mode 100644
index 0000000..50e073b
--- /dev/null
+++ b/documentation/MAKE
@@ -0,0 +1,31 @@
+
+@@ I was thinking about the way make handles inclusion of missing
+ makefile. The following strategy seems like a better choice:
+ if makefile is missing try to rebuild & read it immediately,
+ without parsing the rest of makefile and without reexecuting
+ make later. This approach has at least two benefits:
+
+ + more efficient (no make re-execution) (this is not very
+ important in most cases, thoug)
+
+ + more deterministic: if some of the code after (missing)
+ included makefile depends on some code from it (for
+ example a function) then things can break (example:
+ configuration of ORB).
+
+ - sequential (cannot build makefile in parallel)
+
+
+@! target/pattern-specific vpath
+
+@@ no `target is up to date' message when commnds are executed
+ but target is not changed.
+
+@@ %.o %.d and include
+
+%% implicit double expansion: now supports target/pattern variables
+ and $-
+
+%% create `examples' directory where I can put md5 and maybe hello.
+
+@! shortest stem