Primarily for the following fixes:
4dfc0dd Nodejs: Do not use UV_RUN_DEFAULT during InitPostScript()
ade764e Nodejs: Unset kNoInitializeCppgc for Node 20.6 or higher
a38df85 Fix type check of parameters array with zeek.event
deddadd Fix type check of parameters array with zeek.invoke
Setting this option to false does not count missing bytes in files towards the
extraction limits, and allows to extract data up to the desired limit,
even when partial files are written.
When missing bytes are encountered, files are now written as sparse
files.
Using this option requires the underlying storage and utilities to support
sparse files.
In the past, we allocated a buffer with zeroes and wrote that with
fwrite. Now, instead we just fseek to the correct offset.
This changes the way in which the file extract limit is counted a bit;
skipped bytes do no longer count against the file size limit.
(cherry picked from commit 5071592e9b7105090a1d9de19689c499070749d4)
If an index expression in a table constructor isn't a list, don't abort with an
assertion failure (on debug builds) or a bad tag check. Instead, mark the
constructor expression erroneous and return gracefully. The following...
global a_table: table[subnet] of string = {
1.2.3.4/24 = "unspecified", # should have been [1.2.3.4/24] = "unspecified",
};
...now yields:
$ zeek ./test.zeek
error in ././test.zeek, line 2: table constructor index is not a list (1.2.3.0/24 = unspecified)
error in ././test.zeek, line 2: type clash in assignment (a_table = table(1.2.3.0/24 = unspecified))
When checking exported Spicy types for collisions with existing Zeek
types we previously would also check whether they collide with names in
global scope, i.e., we didn't provide a `no_global` arg to
`detail::lookup_ID` which defaulted to false (since we also provided a
module name I'd argue that the behavior of that function is confusing
and probably error-prone -- like seen here).
This meant that e.g., a Spicy enum `foo::Direction` (automatically in
implicit Spicy module scope) would be detected to collide with the
existing Zeek `Direction` enum.
With this patch we use the `lookup_ID` API correctly and do not check
against potential collisions with globals anymore since it is not
needed.
Closes#3279.
Setting this option to false does not count missing bytes in files towards the
extraction limits, and allows to extract data up to the desired limit,
even when partial files are written.
When missing bytes are encountered, files are now written as sparse
files.
Using this option requires the underlying storage and utilities to support
sparse files.
(cherry picked from commit afa6f3a0d3b8db1ec5b5e82d26225504c2891089)
OSS Fuzz generated a CWD request and reply followed by very many EPRT
requests. This caused Zeek to re-log the CWD request and invoke `build_url_ftp()`
over and over again resulting in long processing times.
Avoid this scenario by not logging commands that aren't pending anymore.
(cherry picked from commit b05dd31667ff634ec7d017f09d122f05878fdf65)
A call to `extract_filename_from_content_disposition()` is only
efficient if the string is guaranteed to contain the pattern that
is removed by `sub()`. Due to missing brackets around the `[:blank:]`
character class, an overly long string (756kb) ending in
"Type:dtanameaa=" matched the wrong pattern causing `sub()` to
exhibit quadratic runtime. Besides that, we may have potentially
extracted wrong information from a crafted header value.
(cherry picked from commit 6d385b1ca724a10444865e4ad38a58b31a2e2288)
* origin/topic/awelzel/deferred-default-non-const-v4:
CreationInitsOptimizer: Use PreTypedef() instead of PreType()
Fix deferred record initialization
testing/btest: Un-deferred record initalization tests
PreTypedef() does not require the PreID() trampoline and ensures
it's only called for IDs that are types. Also allows dropping
the const_cast<> due to id->GetType() returning a const TypePtr which
is different from a `const Type*`...
Put RecordFieldInit instances into creation_inits during parsing and
determine their deferrability in an InitPostScript step. Any
RecordFieldInits can be deferred are moved into deferred_inits.
Closes#3260
* origin/topic/vern/script-opt-maint.Sep23:
fix for ZAM statement-level profiling (broken by GH-3199)
ZAM fixes for compatibility with GH-3249 changes
-O gen-C++ fixes for compatibility with GH-3249 changes minor -O gen-C++ BTest updates
minor BTest reordering to diminish differences with script optimization
* origin/topic/awelzel/3267-avoid-unnecessary-inits:
ID/Stmt: Introduce INIT_SKIP and use in ForStmt
CompositeHash: Skip record initialization when recovering vals
Currently, loop vars are added to a function scope's inits and
initialized upon entering a function with default values. This
applies to vector, record and table types.
This is unnecessary for variables used in for loops as they are
guaranteed to be initialized while iterating.
Initializing fields of recovered records caused running &default expression
of fields just so that they are re-assigned in the next step with the
recovered fields. The second test case still shows that the loop var
is initialized as well even though that's not needed.
Add tests for iterating over records with &default attributes for both,
tables and vectors.
Fixes#3267
* origin/topic/awelzel/3275-fix-with-python:
ci: update-alternative to have python3 be python3.9
Add OpenSUSE Leap 15.5 to ci
configure/CMakelists: Use Python_EXECUTABLE as hint for interpreter
Tweak so that python3 is correct for the Python_FIND_UNVERSIONED_NAMES
change. python3 on 15.4 was pointing at Python3.6 still and for that
there are no dev headers installed.
A normal user could use --with-python instead, but in CI this is easier.
This also adds 15.4 to the list of distributions that are skipped by
default - let me know if this is not a good idea. I assume that by now
most people will be running 15.5 (15.4 will be EOL in ~3 months).
DNS_Mgr has a GetNextTimeout() implementation that may return 0.0. When
that is the case, its IO source is enqueued as ready with an fd of -1.
This in turn results in Process() being called instead of ProcessFd()
in RunState.cc.
Ensure timeouts behavior is properly handled by actually forwarding
timeout indications to c-ares via DNS_Mgr::Process(). This results
in pending DNS queries for which a timeout happened to actually
timeout (when there's no other connectivity).
* origin/topic/jazoff/gh-3268:
Fix check for emailed notices
Changes: Added a test-case printing email_delay_tokens to compare email vs
non-email notice types. Previously, both notice types would have email
delay tokens at that point in the flow.
