[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gcl-devel] Failed building gcl-2.6.9 on NIX using gcc 6.3.0 on x86-
Re: [Gcl-devel] Failed building gcl-2.6.9 on NIX using gcc 6.3.0 on x86-64
Mon, 02 Apr 2018 12:51:31 -0400
Greetings, and thanks for your note!
Blake McBride <address@hidden> writes:
> On Thu, Feb 22, 2018 at 12:33 PM, Camm Maguire <address@hidden> wrote:
> Greetings, and thanks for your report!
> 2.6.9 is way out of date. 2.6.12 is the last official release, and
> 2.6.13pre/list_order in git is the current release candidate.
> When I do a "git branch -a" I see no tags for
> I do see remote tags:
> Should there be a tags:
I must confess to substantial ignorance concerning git 'best
practices'. list_order was a 'toplevel' branch off of
Version_2_6_13pre, but has since been merged in to the latter. This is
the 2.6.13 release candidate, being tested in debian/unstable packages
Just committed a few changes to master which should restore the build
with latest default gcc settings. The compiler is in many ways much
better, but is way too slow for prime time. The inlining algorithm
needs to be rethought, as there are cases where multiple iterative
provisional processing of the same code grows exponentially.
> I haven't been able to build master since the last email I sent.
> sources are also in the latest Debian package.
> I would strongly encourage not attempting to go back to the old inline
> semantics, as support for the newer version has been live in gcl for a
> long time now. I am also unaware of any breakage of
> -fomit-frame-pointer and SGC.
> By way of commentary, SGC, while still supported, does not seem to yield
> any benefit on modern hardware, and does incur a small cost. This is
> basically due to the fact that if you touch swap at all, its 'game over'
> performance wise.
> Parallelism in the build would be possible in the lsp, cmpnew, and clcs
> subdirs only, to my understanding. The logic of the build is highly
> serialized at the moment.
> Take care,
> Jerry James <address@hidden> writes:
> > On Wed, Feb 21, 2018 at 11:33 PM, George Gaarder <address@hidden> wrote:
> >> I ran `./configure; make` and got thousands of `multiple definition of`
> >> error lines. Here is the full terminal session, including configure
> >> https://paste.ubuntu.com/p/qnwnPjrY7d/ where you can see the errors from
> >> line 1868 on.
> > The semantics of 'inline' changed. You need to add "-std=gnu89" to
> > your CFLAGS, since that version of gcc defaults to a later C standard.
> > Also, if you plan to use SGC, you will also need to add
> > "-fno-omit-frame-pointer" to CFLAGS, since that version of gcc
> > defaults to omitting the frame pointer. Lessons learned while
> > building gcl for Fedora. :-)
> >> Besides, I have another question: At the beginning I used `make -j` to
> >> up and seems it makes `make` failed. Why?
> > Yes, I've had -j disabled for the Fedora build for a long time. I
> > don't know how to fix it.
> > Regards,
> Camm Maguire address@hidden
> "The earth is but one country, and mankind its citizens." -- Baha'u'llah
> Gcl-devel mailing list
Camm Maguire address@hidden
"The earth is but one country, and mankind its citizens." -- Baha'u'llah