case in filenames

[René J. V. Bertin]

Michael Pyne wrote:

> 2. Do not rely on file names that differ only in case (e.g. don't install a
> fileA.foo and FileA.foo and expect to be able to open a precise one of those
> two later)
> 
> As far as temp file names and things like that, it would be incumbent on the
> code generating such file names to ensure that they don't break guideline 2
> above. Maybe there's more that should be covered but I think that would take
> care of a large portion of any problems.

Yes, I think so too. 

An example of a framework that doesn't (as far as I can tell) respect guideline 
2: attica. Just 2 examples:
include/KF5/Attica/Attica/AccountBalance
include/KF5/Attica/attica/accountbalance.h

Depending on the order in which those files are installed on a FS like HFS+ (in 
case-insensitive-but-case-preserving mode), you will end up with Attica/Attica 
or Attica/attica directories; the directory name case changes to reflect the last 
write.
Basic calls like fopen() will succeed regardless of case, but there's a reason I 
show an example concerning header files. Compilers (clang at least) do not simply 
try to open a requested include file, but use a search algorithm to locate the 
file first ... and that algorithm takes case into account. So if 
Attica/AccountBalance is requested it will not be found in 
include/KF5/Attica/attica/AccountBalance.

I've considered filing a bug report for this several times, but never did because 
I fail to see a trivial solution that doesn't break all dependent code. Maybe

include/KF5/Attica/c++/Attica/AccountBalance
and add -IKF5/Attica/c++  ?

R.