I had accidentally regressed support for -gdwarf in 461fb499f3cff9038a427eae120fb34defc9ab38 when I changed the logic to
use the already-mapped exe/dll image instead of loading it from disk. The string table is mapped as all zeroes by the loader,
so if a section header's name is longer than 8 bytes (like the ones generated by -gdwarf), then the name can't be read.
Now, if any section headers require the string table, the file is mapped from disk.
windows: Add NtCreateSection/NtMapViewOfSection/NtUnmapViewOfSection
- Fix unwindFrame using the previous FDE row instead of the current one
- Handle unwinding through noreturn functions
- Add x86-linux getcontext
- Fixup x86_64-linux getcontext not restoring the fp env
- Fix start_addr filtering on x86-windows
- .eh_frame_hdr contains a binary-searchable data structure for finding an FDE. If present, we can use this
section to avoid having to parse the entire FDE/CIE list in the binary, instead only entries that are actually
required for unwinding are read.
- rework the inputs pc-relative pointer decoding to support both already-mapped sections as well as sections
mapped from a file
- store the VirtualMachine on UnwindContext so the allocations can be reused
Some distributions (ie. Ubuntu) have their libc debug
info in separate files. This change allows the stack walking
code to read that debug info.
- add support for reading compressed ELF sections
- support reading the build-id from the elf headers in order to lookup external debug info
- support reading the .gnu_debuglink section to look up external debug info
At the moment, the LLVM IR we generate for this fn is
define internal fastcc void @AstGen.numberLiteral ... {
Entry:
...
%16 = alloca %"fmt.parse_float.decimal.Decimal(f128)", align 8
...
That `Decimal` is huuuge! It stores
pub const max_digits = 11564;
digits: [max_digits]u8,
on the stack.
It comes from `convertSlow` function, which LLVM happily inlined,
despite it being the cold path. Forbid inlining that to not penalize
callers with excessive stack usage.
Backstory: I was looking for needles memcpys in TigerBeetle, and came up
with this copyhound.zig tool for doing just that:
ee67e2ab95/src/copyhound.zig
Got curious, run it on the Zig's own code base, and looked at some of
the worst offenders.
List of worst offenders:
warning: crypto.kyber_d00.Kyber.SecretKey.decaps: 7776 bytes memcpy
warning: crypto.ff.Modulus.powPublic: 8160 bytes memcpy
warning: AstGen.numberLiteral: 11584 bytes memcpy
warning: crypto.tls.Client.init__anon_133566: 13984 bytes memcpy
warning: http.Client.connectUnproxied: 16896 bytes memcpy
warning: crypto.tls.Client.init__anon_133566: 16904 bytes memcpy
warning: objcopy.ElfFileHelper.tryCompressSection: 32768 bytes memcpy
Note from Andrew: I removed `noinline` from this commit since it should
be enough to set it to be cold.
Instead, we now have a looser helper called `checkContains(...)`
that will match on any occurrence similarly to `std.mem.indexOf()`.
While at it, I have cleaned up other combinators to make the entire
API more consistent, and so:
* `checkStart(phrase)` is now `checkStart()` followed by
`checkExact(phrase)`
* `checkNext(phrase)` if matching exactly is now `checkExact(phrase)`
* `checkNext(phrase)` if matching loosely is now `checkContains(phrase)`
* `checkNext(phrase)` if matching exactly with var extractors is now
`checkExtract(phrase)`
Finally, `ElfDumper` is now dumping contents of `.symtab` and `.dynsym`
symbol tables. I have also removed dumping of symtabs as optional - they
are now always dumped which cleaned up the implementation even more.
It is assumed that generating a collision requires more than 2^156
ciphertext modifications. This is plenty enough for any practical
purposes, but it hasn't been proven to be >= 2^256.
Be consistent and conservative here; just claim the same security
as the other variants.
