--- Begin Message ---
- To: Debian Bug Tracking System <submit@bugs.debian.org>
- Subject: xvfb: xvfb-run default screen size is unrealistic
- From: John Kozak <jk@thameslighter.net>
- Date: Sat, 12 Aug 2017 16:20:39 +0100
- Message-id: <ygeefsgn77c.fsf@renn.unassigned-domain>
Package: xvfb
Version: 2:1.19.3-2
Severity: wishlist
Dear Maintainer,
The default screen size is set by this line:
XVFBARGS="-screen 0 640x480x8"
This is really unrepresentative of contemporary and causes problems when
running e.g. selenium. Maybe 1024x768x24 would be more realistic?
-- System Information:
Debian Release: buster/sid
APT prefers testing
APT policy: (500, 'testing')
Architecture: amd64 (x86_64)
Foreign Architectures: i386
Kernel: Linux 4.11.0-1-amd64 (SMP w/16 CPU cores)
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8), LANGUAGE=en_GB:en (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash
Init: systemd (via /run/systemd/system)
Versions of packages xvfb depends on:
ii libaudit1 1:2.7.7-1+b2
ii libbsd0 0.8.6-1
ii libc6 2.24-12
ii libgcrypt20 1.7.8-2
ii libgl1-mesa-glx [libgl1] 13.0.6-1+b2
ii libpixman-1-0 0.34.0-1
ii libselinux1 2.6-3+b2
ii libsystemd0 234-2
ii libxau6 1:1.0.8-1+b2
ii libxdmcp6 1:1.1.2-3
ii libxfont2 1:2.0.1-3
ii libxshmfence1 1.2-1+b2
ii xserver-common 2:1.19.3-2
Versions of packages xvfb recommends:
ii xauth 1:1.0.9-1+b2
xvfb suggests no packages.
-- no debconf information
--- End Message ---
--- Begin Message ---
- To: jk@thameslighter.net, 871925-done@bugs.debian.org
- Cc: Branden Robinson <g.branden.robinson@gmail.com>
- Subject: Re: Bug#871925: xvfb: xvfb-run default screen size is unrealistic
- From: Andreas Boll <andreas.boll.dev@gmail.com>
- Date: Wed, 10 Oct 2018 18:21:21 +0200
- Message-id: <20181010162121.gkffe7efnbmrblj3@localhost.localdomain>
- In-reply-to: <yged180mvxr.fsf@renn.unassigned-domain>
- References: <ygeefsgn77c.fsf@renn.unassigned-domain> <CAN4uE+pfZHU8+Mmn+FXeefqhKNGK2LW1DH8dxK-cM74B-_JBRQ@mail.gmail.com> <ygeefsgn77c.fsf@renn.unassigned-domain> <yged180mvxr.fsf@renn.unassigned-domain>
Version: xorg-server 2:1.20.1-3
On Sat, Aug 12, 2017 at 08:24:00PM +0100, John Kozak wrote:
>
> Branden Robinson writes:
>
> > Hi John,
> >
> > On Aug 12, 2017 12:03, "John Kozak" <jk@thameslighter.net> wrote:
> >
> > Package: xvfb
> > Version: 2:1.19.3-2
> > Severity: wishlist
> >
> > Dear Maintainer,
> >
> > The default screen size is set by this line:
> >
> > XVFBARGS="-screen 0 640x480x8"
> >
> > This is really unrepresentative of contemporary and causes problems when
> > running e.g. selenium. Maybe 1024x768x24 would be more realistic?
> >
> >
> > The intention when I wrote this script with Jeff Licquia back in 2001 was
> > not be realistic. 640x480x8 wasn't realistic even then. The intention was
> > fulfill, with as low an overhead as possible, the requirement of a build
> > process on a GUI-less system that nevertheless insisted on having a
> > connection to an X server. Historically, smaller resolutions were really
> > uncommon (512x384 is as about small as you could get and still fit an 80x24
> > character-cell xterm using the "fixed" font without running out of screen
> > real estate). And fewer bitplanes (like 4 "VGA16" or 1 [monochrome]) were
> > possible but also uncommon.
> >
> > What sorts of problems are being caused in Selenium? Can you be specific?
> >
> > Please note two facts:
> > 1. It is straightforward to override this default via a command-line option
> > to xvfb-run. Doing so in a build script is an excellent way to document the
> > stronger requirements of your application.
> >
> > 2. Your recommendation requires about 10 times more RAM to hold the
> > framebuffer than the default, on the assumption that memory access
> > requirements will align those 3 bytes per pixel on a 4-byte boundary.
> > (Without that assumption, 1024x768x24 requires 7 times more RAM.) On 64-bit
> > architectures the alignment requirement may be even worse.
> >
> > Yes, we have more RAM nowadays than we did in 2001, but we also do a lot
> > more things in parallel. Computing tasks expand to consume all the
> > resources available.
> >
> > I don't think it's heretical to change the default, but I think such a
> > change should be considered only after an empirical study of all of
> > xvfb-run's consumers is done. If the default does not serve the majority of
> > cases, I think the argument for changing it is much stronger.
> >
> > Regards,
> > Branden
> >
> > [I'm away from my less-insecure computing environment, so this message is
> > not GPG-signed.]
>
> Many thanks for the thorough reply.
>
> The problem we had was that some selenium tests that had worked suddenly
> started failing after an upgrade of firefox. There FF/Se interface is
> in flux, so we weren't too surprised; but we were surprised that it
> failed only under `xvfb-run`, not when run against a real Xserver or
> when run against a manually created instance of Xvfb (with default
> geometry 1280x1024x8)!
>
> I take your points about resource use and needing to survey users.
> Having said that, I can also see an argument about having two different
> defaults being confusing.
>
> Perhaps the least intrusive solution would be a note in the man page
> about the default screen size?
>
> --
> John
The default screen size of xvfb-run has been changed to 1280x1024x24
in xorg-server (2:1.20.1-3) and the man page has also been updated to
mention the new resolution.
Closing this bug accordingly.
Thanks,
Andreas
--- End Message ---