Re: SH4 build failure - The return of `-lc'

["Yann E. MORIN" <yann dot morin dot 1998 at anciens dot enib dot fr>]

Arnaud, All,

On Wednesday 11 August 2010 002812 Arnaud Lacombe wrote:
> Speaking about that, I'd be interested to have some details about the
> `cc_core_pass_2' step.
> In particular, why do we need libgcc to be built  ?

- NPTL requires TLS (Thread Local Storage) 
- TLS is partly handled by gcc, partly by libc
- gcc handles TLS in libgcc_s
- libgcc_s requires proper crt{1,i,n}.o and dummy lib{c,m}.so for the target

So, gcc depends on lib C, and lib C depends on gcc. Chicken-n-egg problem.

So we have this convoluted sequence:
- kernel headers
- binutils
- cc-core-1
    -> bare cross-compiler with no TLS, no nothing
- libc-headers
    -> headers + start-files, no fancy stuff
- cc-core-2
    -> bare cross-compiler with TLS
- libc
    -> full libc with TLS+NPTL
- cc
    -> full compiler, with multiple languages, TLS...

Old glibc-s with NPTL also required a final step after the full cc, and
once/when we get rid of those old versions, the libc-finish can go.

> There does not seem to be any details about it
> in docs/.

If you can beat me at putting the above into proper documentation,
then that'd be great! :-)

> The 
> closest answer I can get so far is "Merge the NPTL stuff. [...]", from
> May 2007, which is... enlightening ;-)

Well, I hope the commit messages have been improving over time! :-)

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +0/33 662376056 | Software  Designer | \ / CAMPAIGN     |   ^                |
| --==< O_o >==-- '------------.-------:  X  AGAINST      |  /e\  There is no  |
| http://ymorin.is-a-geek.org/ | (*_*) | / \ HTML MAIL    |  """  conspiracy.  |
'------------------------------'-------'------------------'--------------------'

--
For unsubscribe information see http://sourceware.org/lists.html#faq