Re: "stable" version bootstrapping

[Thorsten Glaser <tg at mirbsd dot de>]

Tobias Burnus <burnus <at> net-b.de> writes:

> GCC since 4.8 requires a C++98 compiler, i.e. GCC since 3.4 should be 
> fine. However, who knows when some C++11 features will start to get 

Hrm, indeed.

> used. Thus, why not using the latest compiler which still builds with C, 
> i.e. GCC 4.6 or GCC 4.7. (The latter bootstraps with C++, unless you use 
> --disable-build-poststage1-with-cxx)

Ah, ok, thanks for that bit. Iâll have to see whether they bootstrap
reliably with pcc (did not try that yet, plus Iâll have to write the
appropriate target support for them).

> (My impression that simply different host compilers get used without 
> anyone singling out a specific version. If problems with a host compiler 
> pops up, one can still fix GCC. To do so, it helps if the issue is 
> reported while the version of GCC, which you try to build, is still 
> maintained.)

I think thatâs the key point here: the versions in question are
probably no longer maintained when a problem pops up.

But I kind of like the approach of having a âstableâ version every
once in a while, which can then build all later versions up to and
including the next âstableâ one, possibly with fixes related to that
aim being officially included later even when the âstableâ version
itself is already otherwise closed. It will make compiling almost
arbitrary GCC versions easy, as one doesnât have to build every
in-between version, which is still quite the task.

And in things like BSD ports, the usual way *is* to start from just
a âbase systemâ and compile everything from source, every time, the
availability of some binary packages notwithstanding (these are just
convenience and not the âgoodâ way to do things). Thatâs why this
point is kind of important.

But no need to hurry, Iâm asking a bit in advance, just in case
this needs to get decided first or something.

bye,
//mirabilos