I have a Linux VM where I am trying to compile a simple Go package. The package was retrieved into my user directory with git:
$ git clone [...]/test.go
Cloning into 'test.go'...
done.
$ cd test.go/
$ ls
main.go
I set up the GOPATH and build:
$ export GOPATH=$PWD; echo $GOPATH
/home/vagrant/test.go
$ go build
$ ls
main.go test.go*
So far so good. But now when I try to build again, it fails:
$ go build
can't load package: package .: read /home/vagrant/test.go/test.go: unexpected NUL in input
Deleting the test.go file before building will allow it to build. But this is inconvenient because tools like github.com/codegangsta/gin which try to rebuild the package will fail.
The repository was named [...]/test.go, and the default container directory for git clone is the repo name, so the containing directory is named test.go\.
From go help build:
If the package is main and file names are provided, the file name derives from the first file name mentioned, such as f1 for 'go build f1.go f2.go'; with no files provided ('go build'), the output file name is the base name of the containing directory.
In this case the output is a file called test.go. The problem is:
In the directory containing the package, .go, .c, .h, and .s files are considered part of the package
During a go build if the output from a previous build, test.go, exists, it will be treated as a source file, triggering the 'unexpected NUL in input' message.
The problem can be resolved by renaming the directory to avoid the build output having a name that will be considered part of the package.