test for issue 76

[lvoege]

on a theoretical 64bit Windows machine where allocating 4+GB fails the test would presumably exception out with a bad_alloc. that would make sense as such a theoretical Windows machine would be unable to properly run this test or verify that a fix fixed it.

[mclow]

So what happens when the call to new (allocating 4+GB) fails?

[mclow]

Also, do you have to handle the case when the disk fills up from the write?

[lvoege]

I would say not. when I try the test with a full disk I get an exception and the test fails. as with not enough memory I think it's good that a test that can't usefully verify a fix on some machine fails.

I tried eating those exceptions and at least BOOST_WARN_MESSAGE()ing about it, but I don't actually see those warnings show up anywhere then, so then the test just silently passes when not enough memory or disk. that seems bad.

    {
        temp_file             temp;
        file_descriptor_sink  file(temp.name(), BOOST_IOS::trunc);
        std::streamsize len = static_cast<std::streamsize>(4ULL * 1024 * 1024 * 1024 + 1);
        try {
            std::unique_ptr<char[]> buf(new char[len]);
            memset(&buf[0], 0, len);
            try {
                std::streamsize written = file.write(&buf[0], len);
                std::streamsize pos = file.seek(0, BOOST_IOS::cur);
                BOOST_CHECK_MESSAGE(
                    pos == written,
                    "silent short write"
                );
            } catch (...) {
                if (GetLastError() != ERROR_DISK_FULL)
                    throw;
                BOOST_WARN_MESSAGE(false, "Disk full");
            }
        } catch (std::bad_alloc &) {
            BOOST_WARN_MESSAGE(false, "Not enough memory to run test");
        }
        file.close();
        BOOST_CHECK(!file.is_open());
    }

[jeking3]

Okay, we identified something with a test. Now we need to fix it. :)

[codecov]

Codecov Report

Merging #78 into develop will decrease coverage by 4.31%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           develop      #78      +/-   ##
===========================================
- Coverage    69.04%   64.73%   -4.32%     
===========================================
  Files           80       81       +1     
  Lines         3437     3547     +110     
  Branches      1025      950      -75     
===========================================
- Hits          2373     2296      -77     
- Misses         451      461      +10     
- Partials       613      790     +177

Impacted Files	Coverage Δ
include/boost/iostreams/detail/access_control.hpp	50% <0%> (-50%)	⬇️
include/boost/iostreams/device/file_descriptor.hpp	62.5% <0%> (-37.5%)	⬇️
include/boost/iostreams/detail/double_object.hpp	70% <0%> (-30%)	⬇️
include/boost/iostreams/detail/execute.hpp	73.33% <0%> (-20%)	⬇️
include/boost/iostreams/invert.hpp	41.46% <0%> (-16.44%)	⬇️
include/boost/iostreams/detail/restrict_impl.hpp	57.62% <0%> (-15.53%)	⬇️
.../boost/iostreams/detail/adapter/filter_adapter.hpp	66.66% <0%> (-13.34%)	⬇️
.../iostreams/detail/streambuf/indirect_streambuf.hpp	65% <0%> (-12.78%)	⬇️
include/boost/iostreams/stream.hpp	61.53% <0%> (-11.19%)	⬇️
include/boost/iostreams/read.hpp	78.43% <0%> (-10.25%)	⬇️
... and 52 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 50b4f00...c006eeb. Read the comment docs.