On 14/05/16 13:26, Thomas Morley wrote:
2016-05-14 11:52 GMT+02:00 Trevor Daniels <address@hidden>:
David Kastrup wrote Saturday, May 14, 2016 9:19 AM
James <address@hidden> writes:
It seems that there is some intermittent issue at Savannah.gnu.org
that is stopping me from pulling or fetching (I am getting either
'SSH-message' errors or 'fatal read' errors.
I got it mostly under control by reducing my MTU to 1400 from its
default of 1500 on my wireless interface. It would appear that
something was not dealing gracefully with packet fragmentation and/or
size. While I consider it most likely to have been a problem at my
end,
it is at least conceivable that it might be a problem with the FSF's
network connections. Though I think that Savannah is on quite a
different net than most of the gnu.org stuff, so that's a bit unlikely.
At any rate, just wanted to throw this out for people who might want to
try whether fiddling with the MTU might make any difference with their
setup.
I had a similar problem 6 years ago which was resolved by
changing the MTU value. The details are recorded here:
http://lists.gnu.org/archive/html/lilypond-devel/2010-02/msg00400.html
Just in case these problems are something similar ...
Trevor
Can't get anything here to work.
Tackling MTU doesn't help either.
https://savannah.gnu.org/
states under Latest News:
Savannah suffering networking problems
posted by rwp, Fri 13 May 2016 10:58:25 PM UTC - 0 repliesLast
Friday May 6th Savannah was moved to new hosting in the same
datacenter with many various assorted related and unrelated changes.
Since that time there have been wide spread reports of networking
problems. The FSF admins are aware of the problem and are trying to
resolve it.
Hope they've success soon,
Harm
It seems whatever problem their networking had has now gone away, so I
have been able to at least pull and fetch.
However I cannot get a successful make doc after applying patches.
That is if start with a clean tree and then
autogen.sh --noconfigure
(make my own 'out of tree' build dir)
cd into build dir and then run ../configure --disable-optimising
make
make test-baseline
make doc
That all works.
If I stick a patch in the process I cannot get make doc to complete.
I.e.
autogen.sh --noconfigure
(make my own 'out of tree' build dir)
cd into build dir and then run:
../configure --disable-optimising
make
make test-baseline
cd .. (back into tree)
patch -p1 < diff.file (from Rietveld)
cd back into build dir
make clean (no problems)
make (passes)
make check (passes)
make doc fails and always with the same output (for the last three patches
I have tested for David, Carl and now Thomas Morley)
--snip--
ln -s ./out-www lilypond
mkdir -p ./out-www/lilypond
rm -f ./out-www/lilypond/[a-f0-9][a-f0-9]
cd ./out-www/lilypond &&
/home/james/lilypond-git/build/scripts/build/out/mass-link symbolic .. .
[a-f0-9][a-f0-9]
touch out-www/lilypond.info-images-dir-dep
mv ./out-www/bach-schenker.preview.png out-www/bach-schenker.png
mv ./out-www/bach-bwv610.preview.png out-www/bach-bwv610.png
mv ./out-www/aucun-snippet.preview.png out-www/aucun-snippet.png
mv ./out-www/cary.preview.png out-www/cary.png
make[3]: Leaving directory
'/home/james/lilypond-git/build/Documentation/ly-examples'
/home/james/lilypond-git/./Documentation/GNUmakefile:230: recipe for
target 'out-www/ly-examples' failed
make[2]: *** [out-www/ly-examples] Error 2
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory '/home/james/lilypond-git/build/Documentation'
/home/james/lilypond-git/stepmake/stepmake/generic-targets.make:166:
recipe for target 'WWW-1' failed
make[1]: *** [WWW-1] Error 2
--snip--
sometimes I get variations on a theme and get an additional error line
--snip--
make[1]: Leaving directory '/home/james/lilypond-git/build'
/home/james/lilypond-git/stepmake/stepmake/generic-targets.make:182:
recipe for target 'doc-stage-1' failed
make: *** [doc-stage-1] Error 2
--snip--
Also note this is not just local to one machine but both of the PCs I use
I - quite deliberately- used different distributions (Fedora 23 on one and
Ubuntu 15.10 on another), I also use j5, j6 or j7 depending on the load of
the PC at the time of testing.
Can someone else try to test any of the new patches submitted based on
current master and make sure that make check AND make doc both work (after
test-baseline and a patch has been applied)?
So I cannot with any confidence test patches properly and until I know how
to resolve this, someone else will have to *fully* test patches instead of
me.
Because of all the noise at Savannah I cannot really say when this started
to occur and also remember that both of these PCs run patchy merge which
also does make doc, so logic would state that the issue started *after*
the last merge but I cannot see anything that looks build related edits
(if you see what I mean)?
So unless my build/patch test procedure is flawed (and I have been using
these steps forever, with the additional make clean thrown in after the
patch apply, from David's advice a six or seven months ago)
Any advice, as always, gratefully received.
James
--
--
James
_______________________________________________
lilypond-devel mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/lilypond-devel