Hi! all!
On 2026-08-15 15:03:14, Maciej W. Rozycki wrote:
On Fri, 14 Aug 2026, Bernd Petrovitsch wrote:
[...]
Then why not have one official big header file for the whole standard?
#include <stdlibc.h>
That wouldn't be too unreasonable nowadays I suppose, not IMHO anyway.
The trivial counter example is windows.h (since ages). And at one point
"they" introduced the #define WIN32_LEAN_AND_MEAN (probably) for a
reason ...
I'm not sure what you refer to, is that Microsoft Windows? How would it
Yes.
be relevant to *nix libc?
Because M$FT choose the "you need only one #include for all header files"
strategy and implicitly admitted that it was not the best idea ....
Back in 1970s-80s when the C language standard was being shaped computers
had memories in the range of kilobytes and storage was often floppy disks
with all the performance implications, plus computing power in the low MHz
Above was more in the late 1980ies/1990ies with harddisks ...
range. Nowadays we have memory amounts in the high range of gigabytes to
terabytes and storage that has throughput of gigabytes per second, and
computing power many orders of magnitude better if you factor in not only
the clock rate increase, but all the contemporary architectural features
as well. How could it not matter for the speed of header compilation?
And we may have "pre-compiled header files" though I never found (or more
stumbled upon) a serious page on "how to use them correctly and
efficiently" (though I'm not looking for that often).
LG,
Bernd
--
Bernd Petrovitsch Email : [email protected]
There is NO CLOUD, just other people's computers. - FSFE
mobile: +43 664 4416156 LUGA : http://www.luga.at
