attribute_hidden

Rob uclibc at paypc.com
Sun Nov 27 03:40:21 UTC 2005


Quoting "Peter S. Mazinger" <ps.m at gmx.net>:

> If someone is wondering why all that hassle about hiding internally used
> symbols, some facts now that 100 are hidden

[...]

> glibc-2.3.6 has 23 such relocs (although provides more then uClibc, so
> please understand, that 431 is far away from that (removing these relocs
> will speed up libc)

I realise this sort of question is most likely off-topic to this list, but I
am wondering about this mempcpy/__mempcpy() stuff.  I'm sure SOMEBODY had a
good reason for it, perhaps to ensure one was a libcall instead of just a
macro definition "inline" replacement...  but to my
admittedly-virgin-to-libc-matters eyes, I was surprised to see such an "ugly"
entry point called in a fairly typical userspace library like regex.

By relocs I take it you're referring to "dummy" API entry points which simply
shuffle their params to a "hidden" internal routine to do the work?  Why would
you not just use the "proper" named entry point then for the API-called routine?

Thanks for the interesting information regarding the usage of hidden_attribute
within uclibc.

=R=




More information about the uClibc mailing list