On Sun, Mar 29, 2020 at 01:55:27PM -0500, Bruce Dubbs via blfs-dev wrote:
> On 3/29/20 12:56 PM, Ken Moffat via blfs-dev wrote:
> > On Sun, Mar 29, 2020 at 10:30:39AM -0500, Bruce Dubbs via blfs-dev wrote:
> > > On 3/29/20 10:13 AM, Ken Moffat via blfs-dev wrote:
> > > > On Sat, Mar 28, 2020 at 12:48:38AM +0100, Uwe Düffert via blfs-dev
> > > > wrote:
> > > >
> > > > (changing the subject)
> > > > >
> > > > > Is there any special reason we would want to upgrade llvm but not
> > > > > rustc?
> > > > > Both are rather ugly packages with pretty long pretty similar build
> > > > > times,
> > > > > but recent rustc versions are available way longer than the recent
> > > > > llvm one.
> > > > > I'd even expect rustc to be easier to upgrade due to fewer stuff to
> > > > > test!?
> > > > >
> > > >
> > > > When I built rustc 1.40.0, 1.41.0 and 1.41.1 I used all 8 cores. It
> > > > seemed to me that the install phase was taking a long time, but I
> > > > didn't think too much about that.
> > > >
> > > > For the book, we want timings with 4 cores, so I took the siblings
> > > > offline. A manual ./x.py build took just under 56 minutes. I then
> > > > started the (DESTDIR) install - on 1.39.0 the other day, that took 8
> > > > minutes and I was surprised at how quick it seemed. With 1.42.0
> > > > that took over 57 minutes (longer than the 'build'). Total is 65
> > > > SBU, and the tests took 17 SBU.
> > > >
> > > > The build of 1.42.0 using system llvm was smaller than 1.39.0 with
> > > > its shipped llvm. Oh, and while all of the build, install and test
> > > > were running (for measuring, I do tests afterwards, to see how the
> > > > space changes) from time to time the desktop was unusable, e.g. lost
> > > > a few characters when typing a mail.
> > >
> > > Did you log the install? Is it doing a complete rebuild? What happens if
> > > we just did
> > >
> > > DESTDIR=${PWD}/install python3 ./x.py install
> > >
> > > without the build. I know that the tests would not be run until after the
> > > install. Also, do we need --exclude src/tools/miri on the install line?
> > >
> > > -- Bruce
> >
> > Hi Bruce,
> >
> > I did log it, but after discovering it had been swapping, finally
> > grokking Pierre's comment re the grep and then killing everything
> > unnecessary on that box it is now rerunning the build.
> >
> > Interesting ideas. I'll try to test all of them.
>
> I do not think the above works. I tried and the build failed.
>
> On the other hand, I used the current instructions for 1.42.0 and it took
> about 21 minutes at -j10 for the build and only 2m39s for the install
> portion.
>
> -- Bruce
My repeat build was already running, but when I edited my notes on
what I was doing I found that from time to time the desktop stalled
for a few seconds - top showed 4 rustc processes each running at 103
to 107%.
When the build finished I copied the tree to -built, then went back
in and installed. Nice and quick, although I think it was about 50%
longer than your (I'm back on my laptop, while the build machine
tries a straight ./install with all variables set including
DESTDIR).
The alternative install with --exclude src/tools/miri took an extra
0.38 SBU, but that was probably the normal "rust is a PITA for
varying timings, because it may build in a random order." I noticed
that the sizes in MB for each of those two builds, and the original,
were the same for the build tree, install copied to /tmp, and
~/.cargo : but the total in MB varied by 2MB (first build biggest,
last build smallest).
I'll go back later to see if my "only run the install" also failed.
Thanks.
ĸen
--
When alle is ſayed and all is done, ye must chooſe your faces wisely,
for soon enouff ye will be playing with fyre."
The Nice and Accurate Prophecies of Agnes Nutter, Prophecy 5004
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
