Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Xi Ruoyao]

On 2018-06-20 11:39 -0400, Jean-Marc Pigeon wrote:
> On 06/20/2018 11:22 AM, Bruce Dubbs wrote:
> [...]
> 
> > > 
> > > I'll go ahead and make some changes to do the above and we can comment
> > > on the changes and update that as needed.
> > 
> > I just committed revision 11424.  Please see if that makes the Libstdc++
> > page more informative.
> > 
> >-- Bruce
> 
> Humm.. sorry to ask the question, may be obvious for the others
> but... how to look at 11424?
> URL?, Git?

SVN.  The URL is http://svn.linuxfromscratch.org/LFS/trunk/BOOK/.
-- 
Xi Ruoyao 
School of Aerospace Science and Technology, Xidian University
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

On 06/20/2018 11:22 AM, Bruce Dubbs wrote:
[...]



I'll go ahead and make some changes to do the above and we can comment
on the changes and update that as needed.


I just committed revision 11424.  Please see if that makes the Libstdc++
page more informative.

   -- Bruce

Humm.. sorry to ask the question, may be obvious for the others
but... how to look at 11424?
URL?, Git?


--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Bruce Dubbs]

On 06/20/2018 09:44 AM, Bruce Dubbs wrote:

On 06/20/2018 09:03 AM, Pierre Labastie wrote:

On 20/06/2018 08:38, Xi Ruoyao wrote:

On 2018-06-19 10:27 -0500, Bruce Dubbs wrote:

On 06/19/2018 09:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

Is that really necessary?  The very first line after the title is:

"Libstdc++ is the standard C++ library."
After all, I think "Libstdc++-8.1.0" is quite strange.  In fact the 
version
of libstdc++ is 6.0.25.  Upstream uses the name "GCC 8.1 Standard C++ 
Library"

.

FWIIW: Technically, the installation of libstdc++ completes gcc-pass1. 
It has to be deferred to after building glibc, because libstdc++ uses 
it. Maybe, we could have:


- on the gcc-pass1 page, in "the meaning of the configure options", 
about --disable-libstdcxx (separate it from the other 
"--disable-xxx"): "This switch disables building the standard C++ 
library, which is normally part of GCC installation. The C++ standard 
library is needed to compile C++ code (part of GCC is written in C++), 
but we have to defer its installation, because it depends on glibc, 
which is not yet available in /tools. The installation of GCC-pass1  
will be completed when installing libstdc++ (make an xref to the 
libstdc++ page)"


What we have now on the gcc-pass1 page seems OK to me, but I think it 
would be good to put the above, slightly reworded, into the description 
of the libstdc++ page.


- on the "libstdc++" page: change the title to "GCC- - End of 
Pass1"


I was just going to suggest something similar: "Libstdc++ from gcc-8.1.0"

Change the introduction to: "Now that glibc is available, the 
installation of GCC-pass1 can be completed by building the standard 
C++ library. It is needed for the correct operation of the g++ compiler."


I'll go ahead and make some changes to do the above and we can comment 
on the changes and update that as needed.


I just committed revision 11424.  Please see if that makes the Libstdc++ 
page more informative.


  -- Bruce

--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Bruce Dubbs]

On 06/20/2018 09:03 AM, Pierre Labastie wrote:

On 20/06/2018 08:38, Xi Ruoyao wrote:

On 2018-06-19 10:27 -0500, Bruce Dubbs wrote:

On 06/19/2018 09:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

Is that really necessary?  The very first line after the title is:

"Libstdc++ is the standard C++ library."
After all, I think "Libstdc++-8.1.0" is quite strange.  In fact the 
version
of libstdc++ is 6.0.25.  Upstream uses the name "GCC 8.1 Standard C++ 
Library"

.

FWIIW: Technically, the installation of libstdc++ completes gcc-pass1. 
It has to be deferred to after building glibc, because libstdc++ uses 
it. Maybe, we could have:


- on the gcc-pass1 page, in "the meaning of the configure options", 
about --disable-libstdcxx (separate it from the other "--disable-xxx"): 
"This switch disables building the standard C++ library, which is 
normally part of GCC installation. The C++ standard library is needed to 
compile C++ code (part of GCC is written in C++), but we have to defer 
its installation, because it depends on glibc, which is not yet 
available in /tools. The installation of GCC-pass1  will be completed 
when installing libstdc++ (make an xref to the libstdc++ page)"


What we have now on the gcc-pass1 page seems OK to me, but I think it 
would be good to put the above, slightly reworded, into the description 
of the libstdc++ page.


- on the "libstdc++" page: change the title to "GCC- - End of 
Pass1"


I was just going to suggest something similar: "Libstdc++ from gcc-8.1.0"

Change the introduction to: "Now that glibc is available, the 
installation of GCC-pass1 can be completed by building the standard C++ 
library. It is needed for the correct operation of the g++ compiler."


I'll go ahead and make some changes to do the above and we can comment 
on the changes and update that as needed.


  -- Bruce

--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

On 06/20/2018 10:03 AM, Pierre Labastie wrote:

On 20/06/2018 08:38, Xi Ruoyao wrote:

On 2018-06-19 10:27 -0500, Bruce Dubbs wrote:

On 06/19/2018 09:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

Is that really necessary?  The very first line after the title is:

"Libstdc++ is the standard C++ library."

After all, I think "Libstdc++-8.1.0" is quite strange.  In fact the
version
of libstdc++ is 6.0.25.  Upstream uses the name "GCC 8.1 Standard C++
Library"
.


FWIIW: Technically, the installation of libstdc++ completes gcc-pass1.
It has to be deferred to after building glibc, because libstdc++ uses
it. Maybe, we could have:

- on the gcc-pass1 page, in "the meaning of the configure options",
about --disable-libstdcxx (separate it from the other "--disable-xxx"):
"This switch disables building the standard C++ library, which is
normally part of GCC installation. The C++ standard library is needed to
compile C++ code (part of GCC is written in C++), but we have to defer
its installation, because it depends on glibc, which is not yet
available in /tools. The installation of GCC-pass1  will be completed
when installing libstdc++ (make an xref to the libstdc++ page)"

- on the "libstdc++" page: change the title to "GCC- - End of
Pass1"
Change the introduction to: "Now that glibc is available, the
installation of GCC-pass1 can be completed by building the standard C++
library. It is needed for the correct operation of the g++ compiler."

As usual when I write what I think is English, this text certainly needs
review...


IMHO, Xi Ruoyao was right on the spot with its counter-proposal.
Its consistent with fact, libstdc++ in embedded within GCC and
the way linux header generation was presented in the book.
My 2 cents.


--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Pierre Labastie]

On 20/06/2018 08:38, Xi Ruoyao wrote:

On 2018-06-19 10:27 -0500, Bruce Dubbs wrote:

On 06/19/2018 09:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

Is that really necessary?  The very first line after the title is:

"Libstdc++ is the standard C++ library."

After all, I think "Libstdc++-8.1.0" is quite strange.  In fact the version
of libstdc++ is 6.0.25.  Upstream uses the name "GCC 8.1 Standard C++ Library"
.

FWIIW: Technically, the installation of libstdc++ completes gcc-pass1. 
It has to be deferred to after building glibc, because libstdc++ uses 
it. Maybe, we could have:


- on the gcc-pass1 page, in "the meaning of the configure options", 
about --disable-libstdcxx (separate it from the other "--disable-xxx"): 
"This switch disables building the standard C++ library, which is 
normally part of GCC installation. The C++ standard library is needed to 
compile C++ code (part of GCC is written in C++), but we have to defer 
its installation, because it depends on glibc, which is not yet 
available in /tools. The installation of GCC-pass1  will be completed 
when installing libstdc++ (make an xref to the libstdc++ page)"


- on the "libstdc++" page: change the title to "GCC- - End of 
Pass1"
Change the introduction to: "Now that glibc is available, the 
installation of GCC-pass1 can be completed by building the standard C++ 
library. It is needed for the correct operation of the g++ compiler."


As usual when I write what I think is English, this text certainly needs 
review...


Pierre
Pierre

--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Xi Ruoyao]

On 2018-06-19 10:27 -0500, Bruce Dubbs wrote:
> On 06/19/2018 09:30 AM, Xi Ruoyao wrote:
> > On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:
> > > A)
> > > rename:
> > > Libstdc++-8.1.0 -> GCC-8.1.0-pass2
> > > GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

> > I think we can use "GCC-8.1.0 C++ Runtime Library" just like
> > "Linux-4.17.1 API Headers".
> 
> Is that really necessary?  The very first line after the title is:
> 
> "Libstdc++ is the standard C++ library."

After all, I think "Libstdc++-8.1.0" is quite strange.  In fact the version
of libstdc++ is 6.0.25.  Upstream uses the name "GCC 8.1 Standard C++ Library"
.

-- 
Xi Ruoyao 
School of Aerospace Science and Technology, Xidian University
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Bruce Dubbs]

On 06/19/2018 12:24 PM, Jean-Marc Pigeon wrote:

Hello,

On 06/19/2018 11:27 AM, Bruce Dubbs wrote:
[...]




Do we really need to tell the user how to do this?  It is optional and
the user should really have the knowledge now of how to do this.  I've
never has a student ask how to do it.  They do ask the question "Do I
need to do this?" but not how.

BTW, my answer to the above question is that the conservative approach
is to leave it at this point but they will probably not need it.
Occasionally a package will be missing (e.g. didn't do a make install on
a non critical package) and they may need to go back into chroot using
the command in 6.4.  They are really better off removing /tools at a
later time.  After all it uses about 1.2 GB of space, but that's rarely
an issue with hard disks today.



"Do we really need to tell the user"?
Absolutely yes. We are not telling him how to do it,
I propose we explicitly tell, "Now, you can do it and you must do it"


I disagree with the 'must' part of your statement.


I explain why (from my prospective)
The whole LFS exercise is to build a "Linux from scratch"
/tools directory is not part of Unix/Linux definition
it is lever to do a Linux.

So when you have follow all the instructions you should have
marker saying.
"Now remove /tools, Congratulation you achieve to build
and assemble a limited but fully functional Linux".


You only have a 'fully functional Linux' after you reboot into the newly 
constructed system.



This is a "state change" marker.
As sending a satellite and stating "Orbiting velocity reached",
or when Plane takeoff, stating V1... V2.

Experience:
I removed /tools after "stripping again"
so far so good
But later on I a had a doubt about GCC, so I recompile it and
"make -k check"


You should then use the gcc instructions in BLFS.

  -- Bruce



and this time, check test was a total mess...
What compiler is not able to compile itself?!
Took me a double check to understand runtest was
in /tools/bin and is not avail anymore.

so the question to be asked (no mean intended),
if at the  end of chapter-06, it is not yet a Linux, when
in the "Linux from Scratch" process could we declare
"Bingo!, we have a Linux".



--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

On 06/19/2018 01:35 PM, Ken Moffat wrote:

On Tue, Jun 19, 2018 at 01:24:36PM -0400, Jean-Marc Pigeon wrote:


so the question to be asked (no mean intended),
if at the  end of chapter-06, it is not yet a Linux, when
in the "Linux from Scratch" process could we declare
"Bingo!, we have a Linux".


When the new system has successfully booted ?

In theory, when the kernel (and any desired modules) have been
installed and there is a bootloader entry.  But in both of those
steps, and indeed if people do things wrong anywhere after chapter
6, it is possible to have a system which is not yet usable.

ĸen


This is a quite valid answer.

May be I am wrong, but my answer is/was
when you have a system which is self define (there is
everything in it to rebuild it), booting is just a "detail"
(however important :-})

:-}} Technically, if you need /tools to further
develop you are not yet Linux (unless you say
/tools is mandatory and add a bunch of manuals accordingly).



--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

Bonjour Alain,


On 06/19/2018 01:30 PM, Alain Toussaint wrote:

Experience:
I removed /tools after "stripping again"
so far so good
But later on I a had a doubt about GCC, so I recompile it and
"make -k check"
and this time, check test was a total mess...


You need tcl, expect and dejagnu in the final LFS system (i.e. chapter 6) to
solve the mess...


What compiler is not able to compile itself?!


Checking itself. It probably did compile fine but make -k check is to check
itself. To answer the rest of your message (unquotted), I think runtests is
part of dejagnu.

Might be handy to have the "check" program in your system too.


As stated at begining of this thread.

 TCL, expect and dejagnu should be part
of the chapter-06.

This is the real question.



Alain




--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Ken Moffat]

On Tue, Jun 19, 2018 at 01:24:36PM -0400, Jean-Marc Pigeon wrote:
> 
> so the question to be asked (no mean intended),
> if at the  end of chapter-06, it is not yet a Linux, when
> in the "Linux from Scratch" process could we declare
> "Bingo!, we have a Linux".
> 
When the new system has successfully booted ?

In theory, when the kernel (and any desired modules) have been
installed and there is a bootloader entry.  But in both of those
steps, and indeed if people do things wrong anywhere after chapter
6, it is possible to have a system which is not yet usable.

ĸen
-- 
  Keyboard not found, Press F1 to continue
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Alain Toussaint]

> Experience:
> I removed /tools after "stripping again"
> so far so good
> But later on I a had a doubt about GCC, so I recompile it and
> "make -k check"
> and this time, check test was a total mess...

You need tcl, expect and dejagnu in the final LFS system (i.e. chapter 6) to 
solve the mess...

> What compiler is not able to compile itself?!

Checking itself. It probably did compile fine but make -k check is to check 
itself. To answer the rest of your message (unquotted), I think runtests is 
part of dejagnu.

Might be handy to have the "check" program in your system too.

Alain
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

Hello,

On 06/19/2018 11:27 AM, Bruce Dubbs wrote:
[...]




Do we really need to tell the user how to do this?  It is optional and
the user should really have the knowledge now of how to do this.  I've
never has a student ask how to do it.  They do ask the question "Do I
need to do this?" but not how.

BTW, my answer to the above question is that the conservative approach
is to leave it at this point but they will probably not need it.
Occasionally a package will be missing (e.g. didn't do a make install on
a non critical package) and they may need to go back into chroot using
the command in 6.4.  They are really better off removing /tools at a
later time.  After all it uses about 1.2 GB of space, but that's rarely
an issue with hard disks today.

   -- Bruce


"Do we really need to tell the user"?
Absolutely yes. We are not telling him how to do it,
I propose we explicitly tell, "Now, you can do it and you must do it"

I explain why (from my prospective)
The whole LFS exercise is to build a "Linux from scratch"
/tools directory is not part of Unix/Linux definition
it is lever to do a Linux.

So when you have follow all the instructions you should have
marker saying.
"Now remove /tools, Congratulation you achieve to build
and assemble a limited but fully functional Linux".

This is a "state change" marker.
As sending a satellite and stating "Orbiting velocity reached",
or when Plane takeoff, stating V1... V2.

Experience:
I removed /tools after "stripping again"
so far so good
But later on I a had a doubt about GCC, so I recompile it and
"make -k check"
and this time, check test was a total mess...
What compiler is not able to compile itself?!
Took me a double check to understand runtest was
in /tools/bin and is not avail anymore.

so the question to be asked (no mean intended),
if at the  end of chapter-06, it is not yet a Linux, when
in the "Linux from Scratch" process could we declare
"Bingo!, we have a Linux".


--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Bruce Dubbs]

On 06/19/2018 09:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

There is no libstdc++ package as such (all
other entries use package name), in chapter 6
libstdc++ is generated by a standard GCC build.
I think this suggestion improve understanding about
libstdc++ and GCC intrinsinc link.


In the libstdc++ section we DON'T build GCC.  We only build
libstdc++. This should not be a "GCC pass".


Correct.


I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".


Is that really necessary?  The very first line after the title is:

"Libstdc++ is the standard C++ library."


6. Installing Basic System Software

A)
After MPC and before GCC rebuilding add
entries for:
TCL
expect
dejagnu.

GCC "make -k check" need those packages to do it
and use the one in /tools (generated in chapter 5)
but not yet available from chapter 6 (I think this
not consistent).


"Note that the Tcl package used here is a minimal version needed
to run the LFS tests. For the full package, see the BLFS Tcl procedures."

BLFS: 

Full tcl installation needs to edit {tcl,tdbc,itcl}Config.sh with a lot of
"sed ...".  I don't think we should do that in LFS.


That is correct and the reason we don't build tcl in Chapter 6.


B)
After Xz building, add an entry like
the one "Adjusting the Toolchain" saying
"Removing /tools", to mark the fact the
bootstrap process is now fully completed and
build system is now fully self-sufficient.


No.  The temporary tools would be used in "Stripping Again".
In "Cleaning Up" it's mentioned:

"The reason for this is that the programs in /tools are no longer
needed. For this reason you can delete the /tools directory if so
desired."


Correct.


Maybe we can make this more explicit:

 Now the system is fully self-sufficient.  We can remove the
 temporary tools now:

 rm -rf /tools


Do we really need to tell the user how to do this?  It is optional and 
the user should really have the knowledge now of how to do this.  I've 
never has a student ask how to do it.  They do ask the question "Do I 
need to do this?" but not how.


BTW, my answer to the above question is that the conservative approach 
is to leave it at this point but they will probably not need it. 
Occasionally a package will be missing (e.g. didn't do a make install on 
a non critical package) and they may need to go back into chroot using 
the command in 6.4.  They are really better off removing /tools at a 
later time.  After all it uses about 1.2 GB of space, but that's rarely 
an issue with hard disks today.


  -- Bruce

--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Jean-Marc Pigeon]

Hello,

On 06/19/2018 10:30 AM, Xi Ruoyao wrote:

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:

A)
rename:
Libstdc++-8.1.0 -> GCC-8.1.0-pass2
GCC-8.1.0-pass2  -> GCC-8.1.0-pass3

There is no libstdc++ package as such (all
other entries use package name), in chapter 6
libstdc++ is generated by a standard GCC build.
I think this suggestion improve understanding about
libstdc++ and GCC intrinsinc link.


In the libstdc++ section we DON'T build GCC.  We only build
libstdc++. This should not be a "GCC pass".

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

Agreed, good idea.





6. Installing Basic System Software

A)
After MPC and before GCC rebuilding add
entries for:
TCL
expect
dejagnu.

GCC "make -k check" need those packages to do it
and use the one in /tools (generated in chapter 5)
but not yet available from chapter 6 (I think this
not consistent).


"Note that the Tcl package used here is a minimal version needed
to run the LFS tests. For the full package, see the BLFS Tcl procedures."

BLFS: 

Full tcl installation needs to edit {tcl,tdbc,itcl}Config.sh with a lot of
"sed ...".  I don't think we should do that in LFS.

This is interesting.
IMHO LFS should be a stable starting point.
This mean once /tools removed, we should be
able to redo chapter-6 again without trouble
or missing components (chapter-6 being the very very
minimal stage to a Linux OS duplication process).
So in bootstrap phase (chapter-5) we can have
a striped down tcl. But by the end of chapter-6
we should have fully functional TCL (plus expect, dejagnu).

we (we: stand for all Unix packages contributors) have a
big stratification problem (./configure make us
extremely sensitive to compilation context).
This TCL question may be an example.
Asking to an upstream editor, "what is your packages
level within Unix components layering?" will often give
a puzzling answer.

To come back about TCL, as it is needed by LFS chapter-6,
it should be part of LFS and "fully" functional.
so we should move it/them from BLFS to LFS.






B)
After Xz building, add an entry like
the one "Adjusting the Toolchain" saying
"Removing /tools", to mark the fact the
bootstrap process is now fully completed and
build system is now fully self-sufficient.


No.  The temporary tools would be used in "Stripping Again".
In "Cleaning Up" it's mentioned:

"The reason for this is that the programs in /tools are no longer
needed. For this reason you can delete the /tools directory if so
desired."

Maybe we can make this more explicit:

 Now the system is fully self-sufficient.  We can remove the
 temporary tools now:

 rm -rf /tools


I was not sure doing this after Xz, was late enough.
removing /tools in "Stripping Again" is good enough to
me, it show a clear switch from bootstrap to autonomous mode.






--

A bientôt
===
Jean-Marc PigeonE-Mail: j...@safe.ca
SAFE Inc. Phone: (514) 493-4280
  Clement, 'a kiss solution' to get rid of SPAM (at last)
 Clement' Home base <"http://www.clement.safe.ca;>
===



smime.p7s
Description: S/MIME Cryptographic Signature
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Re: [lfs-dev] Version SVN, chapt 5 and 6 adjustments proposal

[Xi Ruoyao]

On 2018-06-19 09:45 -0400, Jean-Marc Pigeon wrote:
> A)
> rename:
> Libstdc++-8.1.0 -> GCC-8.1.0-pass2
> GCC-8.1.0-pass2  -> GCC-8.1.0-pass3
> 
> There is no libstdc++ package as such (all
> other entries use package name), in chapter 6
> libstdc++ is generated by a standard GCC build.
> I think this suggestion improve understanding about
> libstdc++ and GCC intrinsinc link.

In the libstdc++ section we DON'T build GCC.  We only build
libstdc++. This should not be a "GCC pass".

I think we can use "GCC-8.1.0 C++ Runtime Library" just like
"Linux-4.17.1 API Headers".

> 6. Installing Basic System Software
> 
> A)
> After MPC and before GCC rebuilding add
> entries for:
> TCL
> expect
> dejagnu.
> 
> GCC "make -k check" need those packages to do it
> and use the one in /tools (generated in chapter 5)
> but not yet available from chapter 6 (I think this
> not consistent).

"Note that the Tcl package used here is a minimal version needed
to run the LFS tests. For the full package, see the BLFS Tcl procedures."

BLFS: 

Full tcl installation needs to edit {tcl,tdbc,itcl}Config.sh with a lot of
"sed ...".  I don't think we should do that in LFS.

> B)
> After Xz building, add an entry like
> the one "Adjusting the Toolchain" saying
> "Removing /tools", to mark the fact the
> bootstrap process is now fully completed and
> build system is now fully self-sufficient.

No.  The temporary tools would be used in "Stripping Again".
In "Cleaning Up" it's mentioned:

"The reason for this is that the programs in /tools are no longer
needed. For this reason you can delete the /tools directory if so
desired."

Maybe we can make this more explicit:

Now the system is fully self-sufficient.  We can remove the
temporary tools now:

rm -rf /tools
-- 
Xi Ruoyao 
School of Aerospace Science and Technology, Xidian University
-- 
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page