Hi Robert,
On 2012.01.31 11:27, Robert William Fuller wrote:
Pete,
A flashback from my Windows NT driver writing days: I seem to recall
that MSVC will allow "char str[0]" here. You can have a zero-length
array at the end of a structure as long as the size is explicitly zero.
It's just done with "str[0]" instead of "str[]"?
Yes, MSCV has some partial support for zero-length when they are placed
at the end of a struct, and either [0] or [] can be used (accepts both).
So the MSVC fixes do not remove zero-sized arrays altogether. In fact,
most the of the existing zero-sized arrays from the code have been left
untouched.
The only issue I've had to address are sequences of multiple zero sized
arrays or zero sized arrays at the end of a struct when this struct
itself is used in the middle of another struct, as MSVC doesn't support
this usage.
For the first case, I added an union of multiple zero-sized arrays (eg.
[1], which MSVC has no problem with) and for the second case, an union
with a nonzero element (there's only one place where I had to do that: [2]).
I'm hoping that these changes will cause minimal disruption enough to be
integrated into the next version of the API.
The only annoyance is with having to deal with a string that is not zero
indexed. The other option, to avoid a non zero indexed string would be
to add a struct for the string within the union, with a one byte
padding, but that would add an extra indirection =>
p_iso9660_dir->filename.s.str[0] vs p_iso9660_dir->filename.str[1].
If someone thinks having an extra struct to avoid a non zero indexed
string is the better approach, I can amend the patch.
Regards,
/Pete
[1]
http://git.savannah.gnu.org/gitweb/?p=libcdio.git;a=blob;f=include/cdio/ecma_167.h;hb=4d2c961cf1eaf68090b1dea51c86fc9f9ecf9415#l492
[2]
http://git.savannah.gnu.org/gitweb/?p=libcdio.git;a=blob;f=include/cdio/iso9660.h;hb=4d2c961cf1eaf68090b1dea51c86fc9f9ecf9415#l276
