• X11 1.20 and firefox 75 warning

    From Winston@21:1/5 to All on Sat Apr 25 11:20:48 2020
    What a mess: A couple days spent and still worse off, but at least
    functional again...

    Initially, the system was running X11 1.18 and Firefox 70, both of which
    worked just fine.

    A few days ago, I needed a program for a few minutes so I pkg install'ed
    it. Doing so required upgrading firefox from 70 to 75, but firefox
    upgrades have usually been OK. When I was done, I pkg remove'd the
    program and pkg autoremove'd all the extra stuff it had dragged in. All
    seemed fine so far.

    When next I started firefox, it could only view local files and couldn't retrieve anything from the 'Net. Other programs could still access the
    Net just fine, so it wasn't a DNS or connectivity problem.

    So I figured there must be other stuff that needed to be upgraded, too.
    I pkg remove'd firefox and autoremoved its other stuff, then reinstalled everything. No help.

    So I went whole hog: pkg upgrade [everything, including xorg], but
    without rebooting or quitting X11 1.18 yet. Firefox still wasn't able
    to get anything from the Net.

    So I shut down and rebooted. I then spent the next two days trying to
    get X11 1.20 to work decently.

    The VESA driver which had supported 1600x1200 just fine in 1.18 now
    pretty much refuses to do anything except 640x480. I haven't even been
    able to get it to do 1280x1024 which both the monitor and graphics chip support. According to the log file, even though some of the VESA
    1600x1200 and 1280x1024 modes were marked with a '*' (which I believe
    indicates they're acceptable modes) and even though it describes
    "1600x1200" and "1280x1024" as built-in configurations, it later says:

    Not using mode "1600x1200" (no mode of this name)
    Not using mode "1280x1024" (no mode of this name)

    and ends up at 640x480. Yes, other details probably matter and may be
    causing "no mode", but it was working in 1.18. I at least found a way
    to convince the graphic chip driver to do 1280x1024, so I'm not stuck
    with 640x480, but even that was a struggle.

    X11 gets started up the usual way, using startx which runs .xinitrc,
    just as before. However, even after things have come up, they're not
    quite right:

    * The .xinitrc script calls xmodmap to do the standard (X11 1.20 man
    page listed) commands to swap Caps Lock and Ctrl. It only half
    worked: It left both Caps Lock and Ctrl as Caps Lock.

    I found a workaround: As advertised, running xmodmap again undid the
    swap. Running it a third time succeeded in swapping them. :-/

    * The window manager's rc commands, which included various key bindings,
    were only partially performed.

    Workaround: tell the window manager to re-read its rc file.

    Both of those are key binding issues. Maybe that's significant, IDK.

    * The only good news is that Firefox is now once again able to get pages
    from the 'Net. The bad news is that it spews lot of:

    console.warn: services.settings: Could not determine network status.
    Message: TypeError: gNetworkLinkService is undefined

    as it does so. :-/

    I'm posting this mostly as a warning to everyone else here that this
    particular upgrade may not go smoothly. There may also be other issues
    I haven't encountered yet.

    I plan to read "man xorg.conf" yet again to see if there's something new
    I need to be doing. Adding "Visual 1280 1024" didn't stop the VESA
    driver from going to its 640x480 default, and "Visual 1600 1200" with
    the graphic chip driver ends up creating a pan-and-scan desktop with
    1280x1024 actual, which is interesting, in that it can do that, but
    isn't what I'd like.

    If I figure out anything useful, I'll follow up.
    -WBE

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Martin Neitzel@21:1/5 to All on Sat Apr 25 19:07:19 2020
    * The .xinitrc script calls xmodmap to do the standard (X11 1.20 man
    page listed) commands to swap Caps Lock and Ctrl. It only half
    worked: It left both Caps Lock and Ctrl as Caps Lock.

    Same here since the recent xorg-server pkg (quarterly) upgrade.

    xmodmap from within .xinitrc logs nothing to stdout/stderr but only
    does half its job.

    My work-around for now is not to run xmodmap at all from .xinitrc.
    Instead I run it manually once the first xterm is up. This works just
    fine, as it should.

    I'd really like to know what's going on here.

    Martin Neitzel

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Winston@21:1/5 to All on Sun Apr 26 18:32:29 2020
    I've done more tests.

    * I've tried some BIOS changes and various xorg.config file things, but
    haven't gotten the VESA driver to do anything except its fallback
    640x480.

    * Another new problem is display lag. There's now a lag of anywhere
    from (I'm estimating, since these short times are hard to measure) 100
    - 700 ms for even simple things like scrolling. I'm seeing this not
    only in firefox, emacs, and other applications that manage their own
    windows but also in one of my own applications that I'm certain didn't
    change when I did "pkg upgrade" [all]. Page up/down operations that
    used to be nearly instantaneous now take 200-400 ms to start, and the
    delay seems to be proportional to the amount drawn (e.g., a page full
    of text takes longer than a mostly empty page). In firefox displaying
    a local file (where there's no network or script stuff on the page),
    page up/down now happens piece by piece, the way you'd expect if there
    were some huge load average (there's not) or being done over a slow
    network connection.

    It's hard to know for sure what's at fault, though, since I had to
    switch drivers from VESA to one specific to the built-in graphics chip
    on that system, and I don't have data from 1.18 regarding whether such
    lag would have happened back then with that configuration.

    I've now re-read xorg.conf and tried pretty much all the things I know
    to try, so I guess I'll have to live with it until whatever all got
    broken gets fixed. [Yes, the log indicates that MIT-SHM is enabled.]

    At this point, based on this sample size of one, it looks like unless
    you have to upgrade from xorg 1.18 to 1.20, don't, especially if you use
    the VESA driver.
    -WBE

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Winston@21:1/5 to All on Tue Apr 28 14:09:43 2020
    I previously posted:
    * Another new problem is display lag. There's now a lag of anywhere
    from (I'm estimating, since these short times are hard to measure) 100
    - 700 ms for even simple things like scrolling.
    ...
    It's hard to know for sure what's at fault, though, since I had to
    switch drivers from VESA to one specific to the built-in graphics chip
    on that system, and I don't have data from 1.18 regarding whether such
    lag would have happened back then with that configuration.

    I just ran another test.

    Using the VESA driver running at 640x480 (the only resolution it's been
    willing to run at for me so far), scrolling and the like were the way
    they were under 1.18 and the way they should be: no lag at all. I tried
    a couple different programs and the results were the same.
    -WBE

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Christian Barthel@21:1/5 to Winston on Mon May 4 20:41:04 2020
    Winston <wbe@UBEBLOCK.psr.com.invalid> writes:

    [..]
    Initially, the system was running X11 1.18 and Firefox 70, both of which worked just fine.

    A few days ago, I needed a program for a few minutes so I pkg install'ed
    it. Doing so required upgrading firefox from 70 to 75, but firefox
    upgrades have usually been OK. When I was done, I pkg remove'd the
    program and pkg autoremove'd all the extra stuff it had dragged in. All seemed fine so far.

    When next I started firefox, it could only view local files and couldn't retrieve anything from the 'Net. Other programs could still access the
    Net just fine, so it wasn't a DNS or connectivity problem.
    [..]

    I have had a similar experience once: updated/installed some port
    which updated firefox as well. Firefox crashed constantly
    afterwards. I then installed the `firefox-esr` package which
    worked at that time. After doing a full upgrade a few weeks
    later, the problem was gone.

    --
    Christian Barthel <bch@online.de>

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From John D Groenveld@21:1/5 to bch@online.de on Mon May 4 19:28:35 2020
    In article <875zdbk0fz.fsf@barthel.ch>,
    Christian Barthel <bch@online.de> wrote:
    I have had a similar experience once: updated/installed some port
    which updated firefox as well. Firefox crashed constantly
    afterwards. I then installed the `firefox-esr` package which
    worked at that time. After doing a full upgrade a few weeks
    later, the problem was gone.

    If you are on ZFS root, I recommend creating a backup boot environment
    before invoking pkg(8) upgrade.

    John
    groenveld@acm.org

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Winston@21:1/5 to Christian Barthel on Mon May 4 22:45:33 2020
    In reply to an article I posted, Christian Barthel <bch@online.de> replied:
    I have had a similar experience once: updated/installed some port
    which updated firefox as well. Firefox crashed constantly
    afterwards. I then installed the `firefox-esr` package which
    worked at that time. After doing a full upgrade a few weeks
    later, the problem was gone.

    to which groenveld@acm.org (John D Groenveld) replied:
    If you are on ZFS root, I recommend creating a backup boot environment
    before invoking pkg(8) upgrade.

    When I first saw that things had broken, I considered reverting back to
    undo the changes, but decided against it. I figure I'll have to deal
    with the X11 1.18 to 1.20 and other changes some day anyway, so I may as
    well do so now. Reverting would undoubtedly work, but it'd only be a
    short term work-around.

    OTOH, I agree that it's useful to have the option of doing so.
    -WBE

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)