| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* eval.c (pct_fun_s): New symbol variable, holding
the usr:%fun% symbol.
(fun_macro_env): New static function.
(do_expand): For defun and defmacro, use fun_macro_env
to establish an environment binding the %fun% symbol
macro, and expand everything in that environment.
(eval_init): Intern the %fun% symbol, initializing
pct_fun_s, and also register a global symbol macro in
that name so that we can freely use %fun% everywhere
without worrying that the code will blow up.
E.g. a logging macro can use it to get the function name,
but still be useful in a top-level form outside of
a named function.
* stdlib/struct.tl (sys:meth-lambda): New macro.
(defstruct, defmeth): Use sys:meth-lambda as a replacement
for lambda to set up the %fun% symbol macro. In the :init
case which doesn't use a lambda, an open-coded symacrolet
does the job.
* tests/019/pct-fun.tl: New file.
* tests/019/pct-fun.expected: Likewise.
* txr.1: Documented.
* stdlib/doc-syms.tl: Updated.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The bad situation reproduced as a miscompilation of some prof
forms at *opt-level* 5 or above.
The basic idea is that there is a situation like this
prof t2
... profiled code here producing value in t8
mov t2 t8
end t2
end t2
The code block produces a value in t8, which is copied into
t2, and executes the end instruction. This instruction does not
fall through to the next one but passes control back to the
prof instruction. The prof instruction then stores the result
value, which came from t2, back into the t2 register and
resumes the program at the end t2.
The first bad thing that happens is that the end instructions
get merged together into one basic block. The optimizer then
treats them without regard for the prof instruction, as if
they were a linear sequence. It looks like the register move
mov t2 t8
is wasteful and so it eliminates it, rewriting the end instruction
to:
end t8
end t8
Of course, the second instruction is now wrong because prof is
still producing the result in t2.
To fix this without changing the instruction set, I'm introducing
another pseudo-op that represents end, called xend. This is
similar to jend, except that jend is regarded as an unconditional
branch whereas xend isn't. The special thing about xend is
that a basic block in which it occcurs is marked as non-joinable.
It will not be joined with the following basic block.
* stdlib/asm.tl (xend): New alias opcode for end.
* stdlib/compiler.tl (comp-prof): Use xend to end prof fragment,
rather than plain end.
* stdlib/optimize.tl (basic-block): New slot, nojoin.
If true, block cannot be joined with next one.
(basic-blocks jump-ops): Add xend to list of jump ops,
so that a basic block will terminate on xend.
(basic-blocks link-graph): Set the nojoin flag on a
basic block which contains (and thus ends with) xend.
(basic-blocks local-liveness): Add xend to the case
in def-ref that handles end.
(basic-blocks (peephole, join-blocks)): Refuse to join
blocks marked nojoin.
* tests/019/comp-bugs.tl: New file with miscompiled
test case that was returning 42 instead of (42 0 0 0)
as a result of the wrong register's value being returned.
|
| |
|
|
|
|
| |
* tests/019/load-search.tl: skip a certain test if it is run as
superuser; it fails because superuser is not affected by denied
directory search and execute permissions.
|
| |
|
|
|
|
|
|
|
|
|
| |
* tests/017/str-s.tl: Use (libc) not nil in with-dyn-lib.
* tests/018/forkflush.tl: On Cygwin, produce canned output for first
test case, because the real test case produces some DOS line endings
that cause a mismatch.
* tests/019/load-search.tl: Skip test case involving a directory
with bad permissions being in the load search path.
|
| |
|
|
|
| |
* tests/019/load-search.tl: Add some cases that explore
the load search path.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This test currently fails because when we execute an
unsuffixed file like test/019/a, which exists,
another file is executed instead, like test/019/a.txr.
* tests/019/data/a,
* tests/019/data/a.tl,
* tests/019/data/a.tlo,
* tests/019/data/a.txr
* tests/019/data/b.tl
* tests/019/data/b.tlo
* tests/019/data/b.txr
* tests/019/data/c.tl
* tests/019/data/c.txr
* tests/019/load-search.tl: New files.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This test currently fails. This problem was reported
by Paul Patience, with a repro test case.
The issue is that when compile-file is processing a
(defpackage x ...) form, and the package x already exists, it
fails to recognize the form as a package-manipulating form,
and therefore fails to introduce a "fence" in the output
so that subsequent material is placed into a new top-level
object.
The compiled image fo the (defun foo:fun ()) form in
program.tl causes an error: the foo package does not
exist. This is because the symbol foo:fun is being read
as part of the same object which holds the compiled image of
the defpackage form which defines the package.
It's essentially the same problem as this
(let ()
(defpackage :foo)
foo:bar)
The (defpackage ...) cannot execute until the entire form is
read, but that form contains foo:bar which requires the foo
package to exist.
* tests/019/compile-package.tl: New file.
* tests/019/data/program.tl: Likewise.
|
| |
|
|
|
|
| |
Add three tests; the first and third fail.
* tests/019/load-time.tl: New file.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* eval.c (me_push_after_load, me_pop_after_load): New static
functions.
(eval_init): Register push-after-load and pop-after-load
intrinsic macros.
* tests/019/load-hook.tl: Tests for correct expansion.
* txr.1: Documented.
* stdlib/doc-syms.tl: Updated.
|
|
|
*load-hooks* lets a .txr, .tl or .tlo file specify actions to be taken
when the loading of that file completes, whether normally or via
an exception. They are also honored by process exit.
For instance, with this, we can have a Lisp file that behaves like
a script which cleans up after itself (e.g. removing temporary files)
even if it is not run as a stand-alone program, but invoked
via (load ...). Because it's not a stand-alone program, it cannot
simply use the at-exit-call mechanism. The unwind-protect operator could
be used, but it's inconvenient because it protects a single form.
The *load-hooks* feature in effect protects all the top level forms of a
load, similarly to unwind-protect. Also, unwind-protect does not
guard against a process exit. (However, *load-hooks* does not guard
against an abnormal exit, only normal termination).
* eval.c (load_hooks_s): New symbol variable.
(run_load_hooks): New function.
(run_load_hooks_atexit): New static function.
(load): bind *load-hooks* to nil around load. Implement
the hooks processing via run_load_hooks, taking care to pass the
load-time dynamic environment that has already been undone.
(eval_init): Initialize load_hooks_s and register the *load-hooks*
variable. Register run_load_hooks_atexit with atexit, so the
current value of *load-hooks* is processed on process exit.
* eval.h (load_hooks_s, run_load_hooks): Declared.
* match.c (v_load): Similar changes as in load.
* txr.c (txr_main): Run the load hooks with run_load_hooks immediately
after processing the .txr or .tl file, before entering the listener.
* tests/019/load-hook.tl: New directory and file
* tests/load-hook.tl: New file.
* txr.1: Documented.
* stdlib/doc-syms.tl: Updated.
|
