This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re[2]: modify gcc to handle byteorder issue automatically

Hello Theodore,

TP> I'm using something like the following...
TP> Very crude but may serve as a basis.

Well, probably I need to clarify my question.
It is not a problem to _write_ a code, which will work on any-endian
architecture. However it _is_ a problem when you already have a lot of
written code, and just need to make it running on different-endian
arch. Moreover, sometime it would be wiser not to alter anything in
existing code, but in just a compiler. Not only because a lot of
required work, but because the existing code is stable, working (at
least on some arch), and modification will introduce new bugs,
difficult to fix.

I've thought a lot about it. Note that such change will mean that ANY
word/dword will be put in converted order. Here are some possible

1. Storing a DWORD in void*, and obtaining it via DWORD*. Value will
be converted. union { void * a; dword * b } is an another example.

2. Getting (via dereferencing) any value set by a system call, or
giving a stored dword value to a system call (3rd argument for
accept(), for example). Probably the most simple example is errno.

In my case, only problem 1 is serious. This code doesn't call any
system calls, so it won't be a problem.
So I need to modify a compiler itself. What is the best way to do

Best regards, Tim

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]