Failing to forward free calls to the underlying allocator makes
`ValidationAllocator` unusable for testing allocators while checking for
leaks. This change allows allocators that wrap `std.testing.allocator`
to be tested with `std.heap.testAllocator()` in test decls without
reporting erroneous leaks.
1. When the arena is already empty, resetting with `retain_capacity` no longer
results in allocating a buffer with zero capacity.
This behavior was previously intended by the `(current_capacity == 0)` check,
but wasn't correctly implemented.
2. Resetting with `.{ .retain_with_limit = 0 }` is now equivalent to
`free_all` and a new buffer with zero capacity is no longer created.
This is a useful side-effect of the above fixes.
Previously, when the last buffer in `buffer_list` was retained after
deleting all other buffers, `buffer_list` wasn't updated and pointed
to a deleted buffer.
Previously, the buffer reserved with `retain_with_limit` was missing
space for the `BufNode`.
When the user-provided a limit that was smaller than `@sizeOf(BufNode)`,
`reset` would store a new `BufNode` in an allocation smaller than
`BufNode`, leading to a buffer overrun.
Instead of carrying an optimized version of wyhash in the compiler for
small keys, put it into the std lib where it belongs.
...except it does not match the official test cases. This will need to
be fixed before merging into master. This is an extremely
contributor-friendly task.
Related issue: #15916
I'm seeing a new assertion trip: the call to `enumTagFieldIndex` in the
implementation of `@Type` is attempting to query the field index of an
union's enum tag, but the type of the enum tag value provided is not the
same as the union's tag type. Most likely this is a problem with type
coercion, since values are now typed.
Another problem is that I added some hacks in std.builtin because I
didn't see any convenient way to access them from Sema. That should
definitely be cleaned up before merging this branch.
This is handy if you have a u32 and want a u32 and don't want to take a
detour through many layers of abstraction elsewhere in the std.hash
namespace.
Copied from https://nullprogram.com/blog/2018/07/31/
We need to set rbp last because the arguments are stored on the
stack. If we clobber rbp first, then we will get a segfault when
trying to access the function arguments.
I believe I had already done this with the other syscall* functions,
but not with syscall1, so this allows single argument syscalls like
close to work.
To avoid leaking slots, `fetchRemove` must increment `available` (when the "fetch" succeeds).
Without the `available += 1`, the added test `"std.hash_map repeat fetchRemove"` fails with
run test std-x86-linux-none-Debug: error: thread 432734 panic: integer overflow
.../zig/lib/std/hash_map.zig:1365:28: 0x6471d5 in getOrPutAssumeCapacityAdapted__anon_47495 (test)
self.available -= 1;
^
.../zig/lib/std/hash_map.zig:1308:62: 0x616950 in getOrPutAssumeCapacityContext (test)
const result = self.getOrPutAssumeCapacityAdapted(key, ctx);
^
Alternatively, `fetchRemove` could call `removeByIndex`, though that would entail calling `header()` twice instead of once.
Before this commit, if Walker.next errored with e.g. `error.AccessDenied` and the caller did something like `while (true) { walker.next() catch continue; }`, then the directory that errored with AccessDenied would be continually iterated in each `next` call and error every time with AccessDenied.
After this commit, the directory that errored will be popped off the stack before the error is returned, meaning that in the subsequent `next` call, it won't be retried and the Walker will continue with whatever directories remain on its stack.
For a real example, before this commit, walking `/proc/` on my system would infinitely loop due to repeated AccessDenied errors on the same directory. After this commit, I am able to walk `/proc/` on my system fully (skipping over any directories that are unable to be iterated).
Since we are opening each directory for iteration, we know that we don't need to reset the cursor's directory before iterating. Using `iterateAssumeFirstIteration` skips the cursor resetting which eliminates an `lseek` syscall for every directory opened on non-Windows platforms.
This doesn't seem to actually matter much for performance (1.01 ± 0.02 times faster when walking /home/ on my system) but avoiding unnecessary syscalls is always nice anyway.
crypto.bcrypt: allow very large passwords to be pre-hashed
bcrypt has a slightly annoying limitation: passwords are limited
to 72 characters. In the original implementation, additional characters
are silently ignored.
When they care, applications adopt different strategies to work around
this, in incompatible ways.
Ideally, large passwords should be pre-hashed using a hash function that
hinders GPU attackers, and the hashed function should not be deterministic
in order to defeat shucking attacks.
This change improves the developer experience by adding a very explicit
`silently_truncate_password` option, that can be set to `false` in
order to do that automatically, and consistently across Zig applications.
By default, passwords are still truncated, so this is not a breaking
change.
Add some inline documentation for our beloved autodoc by the way.
Windows defines `recv` and `recvfrom` to return a value of type `int`,
rather than a pointer-sized signed integer, and so should use `c_int`
rather than `isize` for their return types.
Fixes bug causing ReleaseSmall to fail on Windows.
Due to the change in default behavior of ReleaseSmall, debug info are
stripped by default. However because `Compile.create` still defaults to
null, `producesPdbFile` will report true for
`lib/std/Build/Step/InstallArtifact.zig` causing it to fail on copying a
file that does not exist. This commit change the default of strip
depending on `optimize`.
