Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20110921211910.GF5865@port70.net>
Date: Wed, 21 Sep 2011 23:19:10 +0200
From: Szabolcs Nagy <nsz@...t70.net>
To: musl@...ts.openwall.com
Subject: Re: New INSTALL guide, request for feedback

* Rich Felker <dalias@...ifal.cx> [2011-09-21 16:03:37 -0400]:
> Feedback would be appreciated!
> 

looks reasonable


sometimes i use musl without installation
- checkout git repo
- edit config.mak, run make
- use the git repo include/ and lib/ paths directly with -isystem, -L
and the -nostd*, crt*.o, -dynamic-linker etc hackery

works well for single file compilations (without dynamic linking
to other libraries) i have a makefile setup for this and just
copy it around when i want to do some quick testing


>   gcc 3.x and 4.x are known to work. pcc and LLVM/clang may work but
>   are untested, and pcc is known to have some bugs.
> 
when building libc.so with other compiler than gcc
the -lgcc in the makefile should be (obviously) changed
(this is the only change needed for pcc that i know of)

(when compiling other code with pcc+musl a -Dalloca=__builtin_alloca is
needed as well, glibc does this define so i guess that's why they don't)

> Correctly-written build systems should not need -D_GNU_SOURCE as part
> of $CC, but many programs do not use feature-test macros correctly and
actually musl does not do fine grain feature testing wrt _BSD_SOURCE
so some correct programs expecting bsd stuff may need _GNU_SOURCE too
(eg quad_t or some xsi functions with bsd origin)

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.