Old behavior: Crash due to VERIFY, unless we're completely and entirely out of
memory (m_available_ranges being empty), in which case it would return -ENOMEM.
New behavior: Return ENOMEM (and don't crash). In the case of nullptr,
also emit a helpful diagnostic.
Note that MAP_FIXED with nullptr is technically okay, but tends to be a sign
that something went wrong.
Also, this should improve mmap performance marginally, as it pulls the check out
of a loop that does not modify any parts of the check.
UE is now self-hosting! Fixes#5709.
However, this still needs some love: "ue UserspaceEmulator true" spits out tons
of error messages, probably false-positives, and takes about 229 seconds to run.
Then again, true-in-ue-in-ue-in-Qemu is three levels of emulation, so no wonder
it takes a long time! :D
Since there is usually no correlation between guest memory-layout and UE memory-layout,
this option does not make any sense. Especially since we provide nullptr.
The auditing code always starts by checking if we're in one of the
ignored code ranges (malloc, free, realloc, syscall, etc.)
To reduce the number of checks needed, we can cache the bounds of
the LibC text segment. This allows us to fast-reject addresses that
cannot possibly be a LibC function.
Accesses in the header (or trailing padding) of a malloc block should
not be associated with any mallocation since only the chunk-sized slots
actually get returned by malloc.
Basically, allow address-to-chunk lookup to fail, and handle such
failures gracefully at call sites.
Fixes#5706.
We don't want to audit accesses into the region *while* we're setting
up malloc tracking for it. Fetching the chunk size from the header
was tripping up the auditing code.
This returns ENOSYS if you are running in the real kernel, and some
other result if you are running in UserspaceEmulator.
There are other ways we could check if we're inside an emulator, but
it seemed easier to just ask. :^)
We now restart the language server transparently if it crashes.
If the language server crashes too frequently (current threshold is
twice within 3 seconds), we give up and will not attempt to restart it
again. HackStudio will still work fine, but features that depend on the
language server will not function.
To support this change we use a new class, 'ServerConnectionWrapper',
that holds the actual ServerConnection and is responsible for restarting
the language-server if it crashes.
Closes#5574.
- FileDB::get() now returns nullptr if the file is not in the FileDB
- Added FileDB::get_or_create_from_filesystem()
- Added FileDB::add() version that receives that file's content as a
parameter
This fixes an issue were "Find in Files" would not use the up-to-date
content of a file with unsaved changes.
The issue existed because 'FindInFilesWidget' uses
Project::for_each_text_file, which retrieves files by their absolute
path. However, when a file is opened in an Editor, it is created with
a relative path.
This caused us to store two ProjectFile objects that refer to the same
file - one with a relative path and one with an absolute path.
HackStudio keeps a map that stores the different ServerConnection
instances we have open.
Previously, that map was indexed by a project's root path.
This did not make much sense because we only support opening a single
project with each instance of the HackStudio program.
We now index the different ServerConnections by the language name, which
allows us to support talking to multiple language-servers in the same
project (e.g C++ and Shell).
This also fixes an issue where if you first opened a Shell file, and
then a C++ file in the same project, then C++ language-server features
would not work.
Add Bitmap::view() and forward most of the calls to BitmapView since
the code was identical.
Bitmap is now primarily concerned with its dynamically allocated
backing store and BitmapView deals with the rest.
This makes them available for use by other language servers.
Also as a bonus, update the Shell language server to discover some
symbols and add go-to-definition functionality :^)
This is a little bit messy, but basically if an ELF object is non-PIE,
we have to account for the executable mapping being at a hard-coded
offset and subtract that when doing symbolication.
There's probably a nicer way to solve this, I just hacked this together
so we can see "cc1plus" and friends in profiles. :^)
