SVN does not recursively apply ignore patterns, so after creating a
new directory one should not forget to set its svn:ignore property
(otherwise `svn status` will be quite noisy after a testsuite run).
For new test directories, I just copied the property of an existing
test in the following way:
cd testsuite/tests/basic
svn propget svn:ignore > /tmp/test-ignore
cd ../<NEW-DIR>
svn propset svn:ignore -F /tmp/test-ignore .
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15086 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
Technically: when a C-- let-bound variable is assigned to, update its type
so that it is the lub of the types of the initial value and all values
assigned to it, with the type ordering Int <: Addr.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15078 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
It is important not to assume that String.t and Bytes.t will always
share the same representation. Using Obj.magic to convert between
functions would give a very bad example to users considering
a migration, which are very quick to imitate any moral turpitude found
in the standard library.
An unfortunate consequence of the change is the duplication of
String.concat code; other designs would be possible to share more
implementation details between Bytes and String (eg. defined
factorized operations on both in a shared internal module), but if we
consider that String representation may evolve in the future this
coincidence of implementation is really a temporary coindence rather
than an definitive duplication.
I checked that all the small functions introduced are marked as
inlinable. In the case of coercions like this, we could even have the
compiler recognize eta-expansions of the identity function and turn
them into simple rebindings.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15060 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
It kicks in for interface inference and compiler invocations.
safe_string does also alter the output of inferred types if
short_paths is used (since "bytes" is shorter than "string"),
but this requires a change to the ocamldoc driver and isn't
addressed in this PR.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15053 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
Follow-up to commit 15012: keeping checkbound equations is useless when load equations have been forgotten, as all checkbounds involve a load of a block header. So, simplify things further by emptying the numbering after calls and allocs.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15022 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
PR#6484 and PR#6486: CSE across memory allocations can present the GC with memory roots that are illegal.
Plus: lift the previous restriction that all arithmetic ops and loads have at most one result register, this isn't true for ARM with soft floats.
Plus: code refactoring and more comments.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15013 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
Typo in destroyed_at_alloc and destroyed_at_c_call, D17 was incorrectly claimed to be preserved.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15007 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
ocamlbuild lazily traverses directory to build a vision of the
filesystem tree in which the build happens. During that traversals, it
parses any configuration file `_tags` it encouters. PR#6482 was caused
by the fact that the configuration-parsing code used the relative path
of the _tags file, which was correct at the time of traversal, but
stale at the time the lazy thunk was in fact forced (a Sys.chdir had
occured in between).
The first fix is to detect when relative paths become stale, and use
the correct absolute path in that situation.
The reason why this lazy thunk was only forced after a Sys.chdir is
that, at hygiene time, only the subtrees that are selected for hygiene
checking are forced. It is an unexpected behavior that non-hygienic
subtrees could remain unforced for so long (in particular, the
semantics of parsing a configuration file only much later in the build
process is more than unclear); this commit also removes this behavior
by always forcing the whole traversed subtree just before hygiene.
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@15000 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
PR#6211: in toplevel interactive use, bad interaction between uncaught
exceptions and multiple bindings of the form "let x = a let y = b;;".
git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@14977 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02
Gabriel Scherer
Alain Frisch
Xavier Leroy
Mark Shinwell
Damien Doligez
Jérémie Dimino