Add fun_info/2 #1351
Add fun_info/2 #1351
Conversation
src/libAtomVM/nifs.c
Outdated
| if (UNLIKELY(memory_ensure_free_opt(ctx, TUPLE_SIZE(2), MEMORY_CAN_SHRINK) != MEMORY_GC_OK)) { | ||
| RAISE_ERROR(OUT_OF_MEMORY_ATOM); | ||
| } |
There was a problem hiding this comment.
I feel like there need to be better system for using those. GC invalidates any pointers to data on process' heap and some functions use it internally. Not only it makes several GC runs but also prevents allocating terms one by one.
Initially, I implemented both /1 and /2 as NIF but experienced problems of invalid pointers when I tried to turn a cstring into a list – it calls a GC so creating second list invalidates pointer to the first one.
I assume the correct way is to put the first one as a root but then the cstring -> list function needs to take two another parameters.
It's a hard problem but I thought about passing some struct that accumulates how much memory is needed and does it as one pass, something like adding an out parameter GCTracker* tracker that is used before final allocation. This necessarily splits memory calculation and memory allocation.
jakub-gonet
force-pushed
the
jgonet/fun-info-2
branch
from
f054dfc to
6ee08f3
Compare
|
Sorry for straining CI so much. Function name representation changed frequently in Erlang (OTP 21, 24, 25 has all different reprs). |
|
2 side comments: I would rebase this on top of main branch, we are adding new features on that branch now. Also I think this change is worth being mentioned in the changelog. |
jakub-gonet
force-pushed
the
jgonet/fun-info-2
branch
from
6ee08f3 to
1522549
Compare
|
Done. |
src/libAtomVM/nifs.c
Outdated
| case MODULE_ATOM: { | ||
| const term *boxed_value = term_to_const_term_ptr(fun); | ||
| if (term_is_external_fun(fun)) { | ||
| term module_atom = boxed_value[1]; |
There was a problem hiding this comment.
I suggest adding some inline functions to term.h (or use existing ones) in order to access fun inner structure, so we don't leak outside of term.h the fun layout.
There was a problem hiding this comment.
Done, I used single function for that with null checks for arguments. If this adds too much branching, I can split them into three functions.
Also passing entire context, not sure if it should take global and heap separately instead.
jakub-gonet
force-pushed
the
jgonet/fun-info-2
branch
from
1522549 to
da4070f
Compare
Signed-off-by: Jakub Gonet <[email protected]>
jakub-gonet
force-pushed
the
jgonet/fun-info-2
branch
from
da4070f to
3fdcd0d
Compare
|
I noticed that I only preserved |
|
@bettio do you want me to rebase on latest main to potentially fix ESP CI problems and resolve conflicts? |
These changes are made under both the "Apache 2.0" and the "GNU Lesser General
Public License 2.1 or later" license terms (dual license).
SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later