Most notably, x/mod now includes the GOPRIVATE pattern-matching API we
were copying before, so we can use it directly.
Also bump the Go version requirement to 1.15, in preparation for the
import path obfuscation PR, and don't let the gotip job fail the entire
workflow.
Before this change, obfuscating any package using unsafe.Pointer and
with GOPRIVATE="*" would result in errors like:
undefined: unsafe.ZrMmYd1lg
This is because the type isn't plain Go; it's rather a special type that
gets special treatment from the typechecker and compiler:
type Pointer *ArbitraryType
So, trying to obfuscate the name "unsafe.Pointer" will never work,
because there isn't a real Go type definition we can obfuscate along
with that.
Updates, but does not yet fully fix, #108.
This is important, because it would mean that we would obfuscate
nothing. At best, it would be confusing; at worst, it could mislead
the user into thinking the binary is obfuscated.
Fixes#20.
Updates #108.
We need to call 'go env GOPRIVATE' instead of just using os.Getenv so
that we pick up the value from the new ${CONFIG}/go/env file, written by
'go env -w'.
However, we were calling 'go env' at every process start, including the
often tens or hundreds of compiler calls to build all the dependencies.
Instead, do that only once on the first 'garble build' process, and use
os.Setenv to pass that along to future garble sub-processes.
name old time/op new time/op delta
Build-8 1.81s ± 0% 1.74s ± 4% -3.78% (p=0.030 n=5+6)
name old sys-time/op new sys-time/op delta
Build-8 1.45s ± 2% 1.22s ± 1% -16.07% (p=0.002 n=6+6)
name old user-time/op new user-time/op delta
Build-8 10.9s ± 1% 10.6s ± 1% -2.82% (p=0.004 n=6+5)
Error strings should never be capitalized.
A binsubstr line in one of the tests was duplicate and thus useless.
Remove duplicate or trailing spaces in test scripts.
Finally, add a TODO for an optimization I just spotted.
Fixes #2.
Line numbers are now obfuscated, via `//line` comments.
Filenames are now obfuscated via `//line` comments, instead of changing the actual filename.
New flag `-tiny` to reduce the binary size, at the cost of reversibility.
Fixes#93.
The second typecheck lead to the creation of different type objects,
which didn't match the types in the blacklist anymore.
It turns out we don't need the second typecheck,
therfore it is now removed.
First, our original append line was completely ineffective; we never
used that "flags" slice again. Second, we only attempted to use the flag
when we obfuscated a package.
In fact, we never care about debugging information here, so for any
package we compile, we can add "-dwarf=false". At the moment, we compile
all packages, even if they aren't to be obfuscated, due to the lack of
access to the build cache.
As such, we save a significant amount of work. The numbers below were
obtained on a quiet machine with "go test -bench=. -benchtime=10x", six
times before and after the change.
name old time/op new time/op delta
Build-8 2.06s ± 4% 1.87s ± 2% -9.21% (p=0.002 n=6+6)
name old sys-time/op new sys-time/op delta
Build-8 1.51s ± 2% 1.46s ± 1% -3.12% (p=0.004 n=6+5)
name old user-time/op new user-time/op delta
Build-8 11.9s ± 2% 10.8s ± 1% -8.71% (p=0.002 n=6+6)
While at it, only do CI builds on pushes and PRs to the master branch,
so that my PRs created from the same repo don't trigger duplicate
builds.
If the flags list included ["-o" "binary"], we would properly skip "-o",
but we wouldn't skip "binary".
Thus, 'go list' would receive "binary" as the first argument, and assume
that's the first parameter and the end of the flags.
And add a unit test case.
Fixes#82, again.
Otherwise any build flags like -tags won't be used, and we might easily
end up with errors or incorrect packages.
The common case with -tags is covered by one of the integration test
scripts. On top of that, we add a table-driven unit test to cover all
edge cases, since there are many we can do quickly in a unit test.
Fixes#82.
We don't really care about tools other than "compile" and "link". Stop
trying to keep a complete list.
Use "if err := f(); err != nil {" where it makes sense.
Simplify some declarations, and use a better variable name than "fW".
Instead of doing a 'go list' call every time we need to fetch a
dependency's export file, we now do a single 'go list' call before the
build begins. With the '-deps' flag, it gives us all the dependency
packages recursively.
We store that data in the gob format in a temporary file, and share it
with the future garble sub-processes via an env var.
This required lazy parsing of flags for the 'build' and 'test' commands,
since now we need to run 'go list' with the same package pattern
arguments.
Fixes#63.
The following identifiers are now skipped,
because they never show up in the binary:
- constant identifiers
- identifiers of local variables
(includes function params and named returns)
- identifiers of local types
Implement a literal obfuscator interface,
to allow the easy addition of new encodings.
Add literal obfuscation for byte literals.
Choose a random obfuscator on literal obfuscation,
useful when multiple obfuscators are implemented.
Fixes#62
Injected functions were mistaken for functions implemented outside go.
Asm functions:
obj.Scope().Pos() == 0
obj.Scope().End() == 0
Injected functions:
obj.Scope().Pos() == 0
obj.Scope().End() == 1
We now check for the End instead of the Pos.
This requires a bit of extra magic to replace one constant in
runtime/internal/sys, but that was simple enough given that we can reuse
a lot of the code to parse the files and write them to a temporary dir.
We can also drop the -X flags, as runtime.buildVersion is based on the
constant that we replace here.
Fixes#44, again.
Generating DWARF in the compiler object files could take as much as 10%
extra CPU time, while we ignore it entirely at the link stage.
Speeds up 'go test -short' from ~19.5s to ~18.5s on my laptop.
Whilst it may not be particularly common, it is legal to embed fields
where the type has universe scope (e.g. int, error, etc). This can
cause a panic in 2 difference places:
- When embedding `error`, a named type is resolved but the package is
nil. The call to `pkg.Name()` results in a panic
- When embedding a basic type such as `int`, no named type is resolved
at all. The call to `namedType(obj.Type()).Obj()` results in a panic
I'm assuming it is OK to return early when a named type cannot be
resolved.. we could let it continue but I think `pkg` should be set to
nil to be correct, so it'd end up returning straight away anyway.
In the added test, the unexported field used to be garbled.
Reflection can only reach exported methods, exported fields, and
unexported fields. Exported methods and fields are currently never
garbled, so unexported fields was the only missing piece.
Avoiding a type switch for the entire node prevents an indentation
level.
We can obtain obj and pkg early, and return early as well if either is
uninteresting. That means less nil checks later on, which means even
less indentation and complexity.
Also remove the -toolexec equivalent, as it's becoming longer now that
we have GARBLE_DIR, and it might become out of date in the future again.
We don't want users to assume it will work forever.
Carefully select a default that will do the right thing when inside a
module, as well as when building ad-hoc packages.
This means we no longer need to look at the compiler's -std flag, which
is nice.
Also replace foo.com/ with test/, as per golang/go#37641.
Fixes#7.
Daniel Martí
pagran
lu4p
Zachary Wasserman
Andrew LeFevre
Nicholas Jones