Bazel fails to properly extract some tarballs

[bsilver8192]

Bazel extracts some files from some tarballs as empty which should not be. The problematic files appear to often (always?) be large files which are duplicated in the tarball.

I think I've solved this problem in the past by unpacking and repacking the archive, but that didn't work last time I tried it (I think I was using an older tar when it worked).

An easy workaround is to unpack the tarball and replace the files Bazel doesn't like with symlinks to the duplicates.

For example, using the latest Linaro GCC (translated from .xz to .gz without touching the .tar itself) results in gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc being extracted as empty while using tar on the command line (standard Debian Jessie) shows that as being a large binary.

Command to create the file for testing:

$ wget https://releases.linaro.org/components/toolchain/binaries/latest-5.1/arm-linux-gnueabihf/gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf.tar.xz  -O - | xzcat | gzip > gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf.tar.gz

Extracting it on the command line:

$ wget https://releases.linaro.org/components/toolchain/binaries/latest-5.1/arm-linux-gnueabihf/gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf.tar.xz -O - | tar xJ gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc
[...]
$ ls -l gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc
-rwxr-xr-x 2 brian brian 2918784 Oct 14 18:23 gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc

WORKSPACE:

new_http_archive(
  name = 'gcc_repo',
  # I made the file generated above accessible here on my machine.
  url = 'http://localhost:8000/gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf.tar.gz',
  sha256 = 'cf2393e49680c8b28c8feae55c6d2fac513e5f2d24e4b2640284e122bed57594',
  build_file = 'gcc.BUILD',
)

gcc.BUILD (doesn't matter much):

exports_files(['bin'])

Checking:

$ bazel build @gcc_repo//...
[...]
$ ls -l bazel-out/../../external/gcc_repo/gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc
-rwxr-xr-x 1 brian brian 0 Nov  8 03:43 bazel-out/../../external/gcc_repo/gcc-linaro-5.1-2015.08-x86_64_arm-linux-gnueabihf/bin/arm-linux-gnueabihf-gcc