(Patch by Adrien Nader!)
While trying to come up with the previous commit, I noticed that the
various Windows config files had diverged: all but config/Makefile.mingw
were lacking a '%' in a pattern and this prevents the pattern from matching
anything.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14176 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
Rev 14168 (build: replace ocamlcomp*.sh.) broke "make world" (but
surprisingly, not "world.opt") because the argument to ocamlmklib's -ocamlc
has to be a Windows path, i.e. C:\...\ocamlc. Instead it gave a UNIX one,
i.e. /home/foo/ocaml/ocamlc.
Call cygpath in otherlibs/Makefile.nt to fix that.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14175 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
-C doesn't work on at least openbsd's make so don't use it.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14173 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
The "make clean" rules were relying on the $(ROOTDIR) variable.
However this variable is only defined when ./configure runs. This broke the
usual "make clean ; ./configure && make world.opt":
< use old trunk >
./configure # config/Makefile doesn't define ROOTDIR
make world.opt
< move to new trunk >
make clean # Fails early and doesn't clean all files
./configure # config/Makefile now defines ROOTDIR
make world.opt # This fails because of left-over files
An easy solution was to run "make clean" again after configure but this is
not how everyone does and in particular, this is not how the jenkins build
bot seem to do.
The recipes for "make clean" should never rely on values defined by
configure for this exact reason.
As a simple solution, only run the commands that rely on $(ROOTDIR) if
$(ROOTDIR) is set.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14172 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
On Windows, IFLEXDIR is defined as -I"$(FLEXDIR)". The new
mk_shell_and_ocamlbuild_config script outputs:
echo IFLEXDIR="\"$(IFLEXDIR)\""
However, becauses $(IFLEXDIR) contains quotation marks, this becomes:
echo IFLEXDIR="\"-I"$(FLEXDIR)"\""
This unquotes $(FLEXDIR) and breaks the program.
Simply exclude IFLEXDIR from the variables that are handled by the
script. This is actually what the previous scripts were doing: the
exclude list was .*FLEXDIR and matched both IFLEXDIR and FLEXDIR while
the one I made only had \<FLEXDIR\> in it.
I've tried to handle it but there are at least three competing languages
and quoting rules: makefile (both gnu make and several bsd makes), shell
script and ocaml.
There's no human way to do string processing given the portability
constraints: both gnu make and pmake (and its descendants) have powerful
string processing functions but they're not the same.
The only sane way would be to store the configuration in a more evolved
language that is portable, has arrays and powerful string handling
routines in standard like C or awk. From there it would be possible to
output make, shell script and OCaml code easily.
One day. Maybe.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14171 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
Among other files, myocamlbuild_config.ml is built in this directory.
This file defines many variables, most of which are not used and
therefore trigger this warning which is then turned into an error
through -warn-error A.
build: remove warning 32 (unused variable) in tools.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14169 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
This script was built from ocamlcomp.sh.in through sed and is called
instead of "ocamlc" (for instance).
It makes it possible to switch from "ocamlc" to "ocamlc.opt" without
changing anything in the Makefiles, only calling sed.
I couldn't cleanly make it handle both a compiler for the target and for
the build. Instead I'm replacing it and doing as much as possible
directly in the Makefiles.
I hoped it would reduce the number of shell invocations, which would
speed things up quite a lot on Windows but I still had to have at least
one since it's not possible to update a make variable from inside a make
rule: i.e. it's not possible to do X=a, build a.opt and update X to be
a.opt.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14168 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
These scripts sed config/Makefile in order to create shell script and
ocaml code that define the same values as the Makefile but in the
corresponding language.
The sed stuff is very difficult to understand, very brittle and
impossible to change.
Remove that altogether and use a Makefile to run commands like
[ echo FOO=$(FOO) ]. There is still some sed involved (to get the list
of variables in the config and to put rewrite the lists into a list of
commands) but much less and all the expressions are put into variables
with meaningful names with lots of comments.
For config.sh, mkconfig.sh generated lines like :
if [ -z "${FOO}" ]; then FOO=bar; fi
The new script sets the value without checking anything.
I haven't found a reason to do things differently: if there is anything
to override, the right place to do it is after sourcing the shell
script. In any case, I haven't seen any such use.
The mkconfig.sh file also set "WINDOWS=true/false" but the only use is
in the parallel build-system which is implemented in build/ and uses
ocamlbuild as much as possible, and it's going away so it's useless to
set it too.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14167 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
Several commits ago, I made a change that replaced includes for
"otherlibs/unix" with "otherlibs/$(UNIXLIB)", making it possible to
include "win32unix".
Apparently I forgot to do it for debugger/Makefile or skipped it on
purpose since it only matters for cross-compilation and cross-compiling
the debugger is not supported (yet).
Change it now rather than forget it later.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14164 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
This variable is currently unused and it breaks my new script which
replaces mkmyocamlbuild.sh.
The issue is that the "DO" variables becomes "do" in
myocamlbuild_config.ml (lowercased) and "do" is an OCaml keyword.
Since the variable is not used, simply remove it.
Also remove the DBGO variable which only exists in Makefile.msvc64.
Checking for occurrences of this variable elsewhere, I only found
byterun/Makefile.nt which defines it as DBGO=d.$(O) rather than
dbg.$(O).
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14163 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
(Patch by Adrien Nader!)
An earlier change I had made conditionalized building ocamldoc.
I mixed up my variable names and set "WITH_OCAMLDOC_OPT" instead of
"OCAMLDOC_OPT". Later on, $(OCAMLDOC_OPT) would evaluate to the
empty-string, failing to add "ocamldoc.opt" to the prerequesites of the
"opt.opt" and "world.opt" rules.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14162 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
I plan to add at least documentation and deprecation information to
each flagset, so structuring it as a record is important.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14140 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
The second test currently fails, because the plugin is not build if no
target is passed. This will be fixed shortly.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14137 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
I just got bitten by a weird issue where the test I expected to run
was ignored, and never appeared in test runs. I just forgot the final
`()` parameter to the `test` function, and `foo;;` was perfectly happy
to accept an input of non-unit type. Now using explicits `let () =` to
avoid that issue in the future.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14136 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
Jacques Garrigue
Alain Frisch
Wojciech Meyer
Gabriel Scherer