how to avoid bootstrapping?

how to avoid bootstrapping?

[Patrik Hagglund]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 426 bytes --]

Hi,

I just fetched egcs-1.0, but I can't find any instructions on how to
just _build_ the compiler. I don't want to perform any tests, and that
makes (as far as I can see) the bootstrapping process unnecessary. I
think the time I have to wait for the first build to complete is more
than enough.

I will have a look at the Makefiles to see if it is doable. Any help
is appreciated.

--
Patrik Hägglund, patha@ida.liu.se

Re: how to avoid bootstrapping?

[Jeffrey A Law]

  In message < 199712081525.QAA00012@portofix.ida.liu.se >you write:
  > Hi,
  > 
  > I just fetched egcs-1.0, but I can't find any instructions on how to
  > just _build_ the compiler. I don't want to perform any tests, and that
  > makes (as far as I can see) the bootstrapping process unnecessary. I
  > think the time I have to wait for the first build to complete is more
  > than enough.
No, you still need to bootstrap the compiler.

jeff

Re: how to avoid bootstrapping?

[Patrik Hagglund]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 591 bytes --]

> > I just fetched egcs-1.0, but I can't find any instructions on how to
> > just _build_ the compiler. I don't want to perform any tests, and
> > that makes (as far as I can see) the bootstrapping process
> > unnecessary. I think the time I have to wait for the first build to
> > complete is more than enough.
>  No, you still need to bootstrap the compiler. 
>
>jeff

I don't see why. Do you mean that the Makefiles lack support for this,
or that there is something in egcs itself that have to be bootstrapped
to work properly? Can you elaborate?

--
Patrik Hägglund, patha@ida.liu.se

Re: how to avoid bootstrapping?

[Jeffrey A Law]

  In message < 199712081723.SAA03538@portofix.ida.liu.se >you write:
  > I don't see why. Do you mean that the Makefiles lack support for this,
  > or that there is something in egcs itself that have to be bootstrapped
  > to work properly? Can you elaborate?
Using make bootstrap is critical because:

	* The fortran compiler won't build with a non-gcc compiler.
	"make bootstrap" will deal with this.

	* It is important that the C++ runtime libraries get built
	with the egcs compiler, not the compiler used to build
	egcs.  Again, "make bootstrap" will deal with this.

	* "make bootstrap" is also critical because it provides some
	indication that the compiler is functional.


jeff

Re: how to avoid bootstrapping?

[Patrik Hagglund]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1426 bytes --]

Hi jeff,

Your answer only makes me more convinced: To be able to build egcs
without bootstrapping is only a matter of Makefile programming.

> 	  * The fortran compiler won't build with a non-gcc compiler.
> 	  "make bootstrap" will deal with this.
> 
> 	  * It is important that the C++ runtime libraries get built
> 	  with the egcs compiler, not the compiler used to build
> 	  egcs.  Again, "make bootstrap" will deal with this.

This only states that the Fortran part and the C++ runtime libraries
is dependent on the C(?) compiler. Bootstrapping couldn't be the only
mechanism to ensure this dependency.

> 	  * "make bootstrap" is also critical because it provides some
> 	  indication that the compiler is functional.

Yes. The bootstrapping process is used for testing. Usually the
bootstrapping is done in two steps to make it possible to compare the
output from the compiler built by the native compiler with the output
from the compiler built by the new compiler. This makes a total of
three stages. A fourth stage seams to be present in egcs, but I haven't
figured out why.

Personally, I think that the "price/performance" ratio of this test,
is to high for ordinary users. Especially when the last test always
fails on some platforms (according to the documentation). Also, I
guess that only the C compiler is tested; other front-ends (and
libraries) remains untested.

--
Patrik Hägglund, patha@ida.liu.se

Re: how to avoid bootstrapping?

[Jeffrey A Law]

  In message < 199712091321.OAA28819@portofix.ida.liu.se >you write:
  > This only states that the Fortran part and the C++ runtime libraries
  > is dependent on the C(?) compiler. Bootstrapping couldn't be the only
  > mechanism to ensure this dependency.
No, but it's the one we've choosen as the standard way to build the
compiler.  You're welcome to try other methods, but you'll have to
do some digging on your own.  Look in the toplevel Makefile.in if
you want to investigate other means of building the compiler.

  > Yes. The bootstrapping process is used for testing. Usually the
  > bootstrapping is done in two steps to make it possible to compare the
  > output from the compiler built by the native compiler with the output
  > from the compiler built by the new compiler. This makes a total of
  > three stages. A fourth stage seams to be present in egcs, but I haven't
  > figured out why.
A fourth stage is supported, but by default egcs will 3 stage.

  > Personally, I think that the "price/performance" ratio of this test,
  > is to high for ordinary users. Especially when the last test always
  > fails on some platforms (according to the documentation). Also, I
  > guess that only the C compiler is tested; other front-ends (and
  > libraries) remains untested.
It always fails on hpux10 using HP's assembler.

It tests the C front end, the middle end, optimizers & code generators.

It's not perfect, but it's a whole lot better than nothing.

jeff