Fix for #378, bswap on read on big-endian architectures #432
Conversation
testFutureProofing and testMultiPartFileMixingBasic both use fread(&length,4,f) to get a 4 byte integer value from input file. The value read is not converted from the little endian format to the machine format causing problems (eg. test didn't finish after 24 hours). fixes issue AcademySoftwareFoundation#81
…ndation#81. Signed-off-by: Cary Phillips <cary@ilm.com>
|
You could use intrinsics based on the compiler instead of including platform specific headers. Both (recent versions of) clang and gcc provide |
|
The file Cary added, bswap32.h, selects the built in intrinsics you haved noted. It's good to double check these things though, thanks. |
|
I did not want claim that the mentioned builtins were 'different' or 'better'. It was a response to the question he posted in his first comment, i.e. I believe a 5-line The future is bright, though: http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2018/p1272r0.html ;) (reference implementation: https://github.com/slurps-mad-rips/byteswap/blob/master/include/ixm/byteswap.hpp) |
|
I'm going to leave this as is for now, if a simpler method comes along that works on the targeted systems without so many #if's, we can simplify it then. |
Every system seems to have a different implementation of bswap, is there a simpler way than this?
I don't have big-endian machine to test this, so just assuming this works.
Fixes #378