On 4/4/19 4:43 PM, Bruce Dubbs via lfs-dev wrote:
On 4/4/19 4:11 PM, Pierre Labastie via lfs-dev wrote:
I've been testing several things for glibc in chapter 6 recently:
- glibc-2.29 has introduced a hard dependency on Python 3. Previous
versions
were only dependent on perl. So I wondered whether perl was still a
dependency. It turns out that glibc can be built without perl, but
that some
(24 exactly) tests are marked FAIL, because they run "mtrace", which
is a perl
script. Note that even if we do not care about those failures, perl
must be
built in chapter 5 because other packages need it: the first one I've
found in
our build order is "bc".
- the current build method in SVN is good for preventing debug
information
from beginning with /tools. /tools is still mentioned at a few
places, because
the gcc libraries are statically linked, and they have been compiled
without
the option -ffile-prefix-map. This could be prevented too, but it is
not worth
the hassle.
- while testing, I found that the current build method _uses_ the
kernel API
headers from /tools/include, while it should use the ones we have just
installed in /usr/include (it's not a problem with debug information
here,
rather a problem with what is used for building the library). For me,
this is
rather bad, since we cannot be sure that the API headers in
/tools/include
have not been modified during chapter 5.
To use the just installed kernel API headers, the switch
"--with-headers=/usr/include" should be passed to configure. I could
add that
to the book instructions, but it seems nobody cares (maybe it was not
at the
right place, or with the right wording, but I think I've written
three times
about that, and the only return I've had was about debug information.)
thank you for the clear explanation. Please go ahead and add the
--with-headers entry. You might want to go ahead and double check the
glibc dependencies in the appendix.
-- Bruce
I apologize for not speaking up earlier.
I think that the --with-headers change is very important. It will remove
the chance of contamination from the host system, and allow us to use
the proper headers.
I vote for proceeding with the changes.
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
