Optional variables are needed because the structure of the makefile
is prepared to run on Guix but Guix doesn't support RISCV yet, so it's
better to set them as optional and let the user decide how do they want
to compile this thing.
* doc/ref/srfi-modules.texi (SRFI-64 Writing Basic Test Suites): Fix
typo. Add default test runner example. Add test-approximate and
test-error examples. Document valid error types in Guile for test-error.
(SRFI-64 Conditonal Test Suites and Other Advanced Features): Fix typo.
Fixes <https://bugs.gnu.org/75041>.
Suggested-by: Arne Babenhauserheide <arne_bab@web.de>
In some situations, ttyname may return ENODEV even though isatty is
true. From ttyname(3):
A process that keeps a file descriptor that refers to a pts(4) device
open when switching to another mount namespace that uses a different
/dev/ptmx instance may still accidentally find that a device path of
the same name for that file descriptor exists. However, this device
path refers to a different device and thus can't be used to access the
device that the file descriptor refers to. Calling ttyname() or
ttyname_r() on the file descriptor in the new mount namespace will
cause these functions to return NULL and set errno to ENODEV.
Observed in a Debian riscv64 porterbox schroot.
When ttyname fails with ENODEV, just include the file descriptor integer
value instead. Call ttyname() directly to avoid having to catch the
ENODEV.
* libguile/fports.c (fport_print): fall back to the integer fd when
ttyname() fails with ENODEV.
Only hold scm_i_misc_mutex while making the C calls. This also avoids
the need for a dynamic-wind. Add SCM_I_LOCKED_SYSCALL (similar to
SCM_SYSCALL) to handle the locking and EINTR loop.
libguile/filesys.c (scm_readdir): rely on SCM_I_LOCKED_SYSCALL to limit
locking.
libguile/filesys.c (scm_ttyname): rely on SCM_I_LOCKED_SYSCALL to limit
locking.
libguile/syscalls.h: add SCM_I_LOCKED_SYSCALL.
* module/language/tree-il/fix-letrec.scm (compute-sccs): Instead of
depending on all previous complex bindings, we can just depend on the
most recent one. Decreases the graph size.
We were using list sets, which when you end up with thousands of
bindings in an SCC reaches the point where we are off the quadratic end
of the curve. Fix to use intsets and intmaps instead.
* module/language/tree-il/fix-letrec.scm (compute-ids): New function.
(compute-referenced-and-assigned): Rename from analyze-lexicals, and
compute intsets.
(make-compute-free-variables): Rename from free-variables, return a
procedure instead of a hash table, and use intsets. Use a global cache
to avoid quadratic behavior with regard to binding depth.
(compute-complex): Compute a global set of "complex" variables, as an
intset.
(compute-sccs): Use intsets and intmaps to compute the free-variable and
ordering edges.
(fix-scc, fix-term): Refactorings.
(reorder-bindings): Avoid a linear search.
(fix-letrec): Refactor.
Releasing memory proceeds until there is (-NOFL_BLOCK_SIZE,0] bytes to
release; we should only expand when the number of bytes to reacquire is
large enough.
This fixes an issue in which minor collection of a nursery full of live
data can fail because of fragmentation, whereas really it should just
fall back to promotion.
If two tracer threads visit edges to the same object on an evacuation
candidate block, and they first see that the object is unmarked, then
they both try to evacuate it at the same time. Thread A might try and
fail before thread B manages to acquire the forwarding word. B needs to
see that A marked it in place, and to do that it needs to re-load the
mark byte after acquiring the forwarding word. Otherwise perhaps B
could succeed and you would end up with two copies of an object, one of
them garbled!
This allows field logging to proceed during collection, which might add
new edge buffers. Also fix a bug that would cause debug-mode assertion
failures, where clearing a field set didn't clear edge buffer next
pointers.
