Welcome to Totally Built Linux Distro
new pre-releases: 12/17/2024 src11+wifi 5/15/2025 src16 7/14/2025 src17
11/17/2025 src17
2025-News displays README.txt containing news for folder 2024/
News is a summary of features added in 2024-2025
. first release 2022 ended in src9 ff109 with USB
. 2025 release ended in src16 USB and has recent changes to src16 still
(the source-additional-src16.txt will likely not change for a while)
(source-additional.txt will combine 14 16 additional)
. less clutter in 2024/README.txt for install/use minded instruction
-> src17 is "the latest stable"
runs firefox-142 Mathematica-14 on nVidia (gtx 1050) or iCore7 UDH 770
NEW src17 11/17/2025
. src17: updates from linux-6.10.2 to linux-6.15.5;
-> unlike src16 it includes games and optionals built
(by mistake - which accounts for the extra 1G download size)
sources-src17-additional.tar does not contain '6.15.5's firmware gz'
(i personally use src16's firmware brief - for shame!)
have not yet used the exact same src17 USB long, however i was using a
near copy since 7/14/2025. just installed new all seems normal
! FLASH ADVISORY of CORRECTIONS !
. 7/14/2025 minor issue of runx flubbing first attempt after power outage
and inefficiency: updated runx is in src16, src17
. there were a few nos/boot or build fail advisories in making of src16
in mid 2025 that can be forgotten now - a big one was source-all.tar
uploaded mistakenly had softlink instead of real tarballs for many items
. there still may be some nos/ quirks it is not totally enduser friendly
most particularly do not press Ctrl-C during first startup (instead
answer question(s))
. i have temporarily lost access to updating this due to AWS script bugs:
https://totally-built-linux-distro.s3.amazonaws.com/index.html
(the web page and uploads are old please dont use)
! FLASH ADVISORY of CORRECTIONS !
news summary 2024 2025
. added several apps (some optional separate download) like blender-3.3.8
apps not everyone uses as to mathematica will be optional downloads
trying to keep free of clutter
. added harware support for iCore7 graphics
. updated X again, which in turn upped some of the build chain
. new nos/ are runtime os files that makes a live USB of tbld
includes optional simple wifi access
. updated build chain tools in src11 and base libs for Xorg-21
. kernel includes basic igc RTL ethernet and icore7 w/IGC EFI graphics
. build in more than one srcN in parallel (not a new feature)
. use of multiple processors during each build make -j N
(fixes already know Makefile that cannot, end-to-end is much faster)
. Better support for making USB: now both Grub and EFI work and on same USB
and are done automatically and it works with more than one kernel made.
. Still builds all older srcN, equal tested to do so and XFree86 build is back
. Easier end to end, though not "one touch" yet. allot of xxx-fixme are
newly fixed so that a build fail is easier to contend with
. NEW: a way to launch building many srcN OS at the same time; see "assay"
. sorry, booting and kernel still not tested for laptop
News
Apps added recently as OPTIONAL (build now has recipe for)
If prefixed '-' it is not in 5thtoe.tar, else it is. All have build recipe.
ALL
-LLVM-19.1.0-Linux-X64.tar.gz
gtk upgrade and Intel GL/multimedia can only build if llvm is present
llvm is Not built (Intel libs should have exact release of llvm)
other optionals may require llvm to build - safe to assume you need it
GAMES
that build from src using tbld and run-tested to open and play
-gemrb-0.9.3 KQ-0.99 (thus allegro dumb lua openal libglade also)
. full graphic games with sound
nethack-3.4.0 -avanor-0.5.8 -omega-rpg-0.90
. nethack is in all.tar, w/ncurses, X11 fonts or X11 graphic tileset,
net not compiled in for security, no sound
. avanor omega have no tiles to enable
infocom xgammon xasteroids xjump xtron xgnuchess xfrisk xscorch
-freedroidRPG-0.16.1 -freedroid-1.0.2
-freeciv-1.14 | 3.1.4
-gearhead-2.0.701 (needs -fpc-3.2.0-x86_64-linux.tar free pascal)
games todo: like to find 1 good true JRPG w/source even if
Qt-install req'd so far only KQ found
OFFICE
-firefox-44 (42 44 cannot browse most sites https htmlver are too low)
not on src17 USB (builds but is usually skipped for time)
-Oracle Java 24 (the Oracle download works, most demos tested)
-qt-5 | -6.8.2 not on src17 USB (was built but not installed)
-blender-3.3.8 not on src17 USB (was built but not installed)
xpaint-2.4.9 xfig-3.2.6 magicfilter-1.2
-readline-8.2 -TiMidity++-2.13.22 cpio-2.8 -wxWidgets-3.2.5
-apache2 not on src17 USB (not installed because apache1 is installed)
-fltk-1.4.1 -tcl8.6.12 -tk8.6.12 -Tix8.4.3 -afterstep-2.2.12
Maxima-5.9.2 | -5.47.0 -wxmaxima-16.12.2
*** see maxima note and wxmaxima note
-Emacs-1.30 TeXmacs-1.0.5 | -1.99.18 (both compiled with gcc-14)
tetex-3.0 gnuplot-5.4.7 clisp-2.49 guile 1.6.8 | -1.8.6 | -2.2
Yacas-1.5.0 -Fricas-1.3.11 ~ -open-axiom-1.4.2
Fortran and MultiProcessor gcc-14.2 recompiled with fortran (is option)
-Octave-3.2.4 (lean) | -10.1.0 with: lapack/blas-3.2|3.12.1
-SuiteSparse-7.10.2 -sundials-7.3.0 fftw-3.3.10 qhull-8.1-alpha1
qrupdate-1.1.2 glpk-5.0 -hdf5-hdf5_1.14.6 arpack-ng-3.9.1
many office tools are in sources-*.txt are built by default not listed
above: see list.*.test: ie: dict rsh apache sendmail bochs xview povray
office todo: toomany. octave needs ?some/a gui, emacs texmacs needs
more testing, jupyter "req's ubu", ...
News
. 11/17/2025 src17 usb uploaded; contains installed optional listed above
one benefit is 6.15.5's console fonts are readable on LED TV
realized parallelport is not in kernel - few will want but B aware
. 7/14/2025 src17 upgrades to linux-6.15.5 (no usb yet), fixed script runx
. 5/15/2025 the build of srcAll with assay data finished, no problems.
while NLS ended up not currently a problem it was in the past before
fixes and ability to "have lean thick" option sets was a long term goal
. src?.gz will be updated with the few updates soon (few days?)
. chroot1.go [no_global_opt_pref] turns off use of new data
. true utf8/16/32 builds take 2x time of ASCII literally so not enabled,
note some apps choke on utf8, gettext runtime can choke some nls
. 5/13/2025 src{9,10,11,12,20,21,16} assay data is in
The impact is 3,148 app builds go with NLS allowed where before disabled
. 5/8/2025 uncovered some fix rebuilding all srcN:
. changed preen-gz-softlinks to accept .bz2 .xz (are in optionals)
. pascal (optionals) fix "question pops up stalling build"
. in src16 only, corrections:
. one stray line in src16/build removed
. optional section: found PATH="path" flub in tetex postinst
. a MANPATH file exist flub
. gearhead missing a "
. in src9,20,21 possible Motif build fail (bash wait) added sleep jic
. 5/5/2025 src16 adding simple feature for build assay to run automated
configure args data:
. build.global.opt has --disable-nls but global.opt.nls does not
. adds --enable-nls if detected (never is)
. if nls build works it records which option assay file worked
. if nls build fails fails it tries again with orig b.g.o.
. data is per srcN per chrootN
. launch-build runs 6 os builds simultaneously so that data will not
effect past builds. also use of the data is optional (ie: always safer
to build w/o nls doc even if it is tested to work)
. 5/5/2025 src10 a mystery change-of-behavior of gcc-10 caused need to
fix gcc-4.7.4 libitm to build by gcc10 (though not used) then disabled
. 5/5/2025 src16 removed feature "nobuildfail" (temporarily?)
. 5/4/2025 src16 minor runx change - check for stale socks and auths
. 5/2/2025 src16 updated 5thoe with any small depends recent optionals need
. 4/16/2025 src16 *** What to expect for TeXmacs with Maxima
texmacs-1.0.5 allows maxima-5 fancy output integrals and to edit those
w/cursor and be copied as text input back to maxima easily.
. 4/27/2025 src16 texmacs-1.99.18 is 12x size of 1.0.5 - idk if the extra
features help maxima-5.47 or not. it will give img load errors for some
help page icons which can be ignored. octave not yet tried fancy mode.
. 4/29/2025 open-axiom may not be safe to use (lisp brand issues),
wxmaxima takes a kick to start (see list.test.3 or build)
wxmaxima does not allow a cursor in fancy output but can copy it as
plain text for re-input (unsure if ubuntu-only 25.x ver does either).
It has allot bugs as things stand it isnt an endpoint for regular use.
25.x CMakeLists.txt works ONLY on ubuntu - tbld cannot build it - and
i would bet it doesn't build on ubu either.
. 4/15/2025 xmaxima (tk) has same ascii art maxima does - it can Not show
fancy fonts nor graphics with text. it has a few useful features
. 4/13/2025 Qt not installed but no policy not to in future "per say"
as size not bad v. gtk: however some apps delete gtk motif x11 sup due to
hacking from for-profit Qt which i feel is illegal, but other apps
ie open-rpg-maker never had gtk - so that is ok
. 4/9/2025 src16 i suggest avoiding SDL-1 games they tend to segfault
and b-needy; port to SDL2 before building could save time and trouble
as for games not listed above
. 4/6/2025 src16 optionals OFFICE+GAMES have been rebuilt back-to-back
with list better organized so it optionals go smoothly
. 4/1/2025 src16 preliminary build of tetex
needs testing to see if it works "as usually expected", like hyphenation
. 3/27/2025 src16 re-upload srcN for minor /etc/cron.dail/find issue
does not fix on USB, /sbin/run-parts should (also) be in /bin
. 3/11/2025 src16 adds as optional and separate dl
. xfig, lp util magicfilter for older Epson LQ, IBM pin, HP ps laser
. gemrb-0.9.3 Baldur's Gate engine & kit (very short demo but good)
(has no full editor, but can run some store games on linux)
. 3/8/2025 src16 adds KQ as optional to build with no-dl req'ed
. game, low 8M footprint, JRPG, sound+music, mapeditor incl., hrs of play
. 3/7/2025 src16 USB preliminary release
the USB was install tested (todo: /var was not tested it pre-existed)
. removed USB 15 (a go-between), choose usb 16 or older 11
. src14 removed (rather keep as is 11 w/llvm, but 11 has usb)
anway not all past scrN will rebuilt just ones good for testing
or that have usb
. 3/6/2025 src16's update (mostly add a few recipe) now tested
compiling each past srcN using same - they all still build
. 3/4/2025 src16 more gtk ff rats to chase, ff ignores gtk4 (has file
dialoge), gtk-3.x is seen, 3.94 removed it and clobbers 3.24's .xml files
... gtk downgraded so "firefox quits less frequently"
. How long did upgrade take from src11 ?
. upgrade from src16 USB hd-hd-install to "Xorg.0.log"
8pm to 8:45pm incl. xfer of kernel over lan and nvidia-installer
. 3/4/2025 src16 TiMidiTy MIDI optional and src20 tested with new ./build
(all srcN will be uploaded when re-tested)
. 3/3/2025 src16 now installed on i5 with nVidia gtx 1050
see In-depth News for src16:
. 3/2/2025 src16 (Intel Quick Sink for ffmpeg)
see In-depth News for src16 for
. what resolution and speeds are now smooth, which still lag
. how to set up QHD TV full screen 1440p that is pleasant viewable
if you have 4k
. 3/1/2025 src16 has been USB tested and as feared Xorg as usual had
startup failure. xf86-input-keyboard-1.8.0 was an auto 1.9,0 works.
(2.1.0 was tried, it is missing symbols)
. 3/2/2025 src16 (Intel Quick Sink for ffmpeg)
. 2/27/2025 src16 builds blender-3.3.8|3.3.4 "lite" (mid 2023)
(no numpy quadri ocean audacity ffmpeg?! serveral other NO)
if src15 is parent os, ran from inside chroot3 without install to system
and could visit all upper tabs and color then render cube. i'm not a
game fanatic myself i hope a lite start is better than nothing.
. dl official blender-3.3.4.tar.gz
. many dl are needed to turn on extra feature back ON
. todo? 3.4 - 4.3 have invasive requirements hard to awk OFF
. 2/27/2025 src16 builds qt-everywhere-src-6.8.2 (45 min?) with not
much work-around. build is optional. dl is separately first,
(sans node.js "google suite", thus QtWebEngine QtPdf, not attempted)
. 2/25/2025 src16 just began compiling with new 2025 Xorg replacements
. in src15 qt-everywhere-src-6.8.2 multimedia forcing upgrade of xorg
that until now hadn't been a problem (2013/2022 to 2025)
. qt-6.8.2 will be recipe only (optional dl)
. next up will recipe for any but latest blender lite, optional,
(there are problems with c++29 custom compiler blender uses, gcc14
is newest but not "understanding jargon" in their code)
. src 10,11,12,14 might be (re)moved from sources-all.tar, 16 was goal
. 2/25/2025 src15 snafu wget had no ssl (curl did) now has 2 bins, one
with ssl one without (incase ssl libs dont load runtime)
. 2/23/2025 src15 modest ads to optionals (these aren't in sources.tar but
if dl/ and uncommented in list have new recipe): apache2 readline-8.2
fltk-1.4.1 tcl8.6.12 tk8.6.12 Tix8.4.3 afterstep-2.2.12 expect5.45.4
(took too long in the preparing, low popcorn, but se la vie)
. 2/14/2025 src15 was installed showed a little Mathematica and Youtube
successes listed further below. It can still be built without these
extras. No more USB14, use USB15, same with optionals added really
. 2/12/2025 src15 src14 rebuild failed so (src15's had nasty libtool issue)
I AM CHANGING POLICY ON UPLOADS
An issue I know well is small changes can cause easy or complicated
problems and I don't like listing them or posting srcN that may have
something nasty to solve. Even if early access might be nice i tried it
and I don't like it and lists of srcN fixes on given dates are confusing.
. 2/7/2025 src14 adds confirmed iCore7 UHD 770 A/V hw accel in firefox.
sources-additional dl needed, src14 USB is not updated.
Youtube 4k fullscreen works not perfect but is accelerated.
Not all video settings (youtube) supply acceleration and note resizing
matters. ex: choosing 4k on a 4k screen is must faster than 1080p.
(rescale X to Y depends on format translation ability sw hw combos)
. nVidia GL A/V setup is separate, 550 setup installs own GL A/V libs
. 1/27 src14 intel-ui rt working, ui was definitely not compat gnu bison
Narkive: tools=intel,intel-ui does not provide a ui despite depends gtk3
Narkive: intel-rt provides no material raytracing entry pt, demo, docs
i will not test raw ".so abi" for rt ui functionality
. 1/26 found live-usb (nos) had a few clerical FLASH ADVISORY
nos/ is new it is tested only on first boot of usb releases,
not nearly as frequent as srcN, though post-install testing is frequent
. 1/11 - 1/24 mesa-24.3 intel(s) now work, you can build swrast only if
you like or can stick with mesa-13 (no LLVM mesa13) if you like
See list.test.3, --__llvm (uncomment line to enable optional build).
The "xorg-damage" mentioned at EOF (mesa-24 llvm intel not mesa-13)
src11 and src15 need uxa and or tearfree setting. src16 does not.
xorg.conf
Section "Device"
Identifier "Card0"
Driver "modesetting"
Option "TearFree" "False"
Option "AccelMethod" "uxa"
BusID "PCI:0:2:0"
EndSection
Xorg -configure should set this up, but if not the "known good" operation is
not evdev or libinput it is Core mode as so:
Section "ServerLayout"
Identifier "X.org Configured"
Screen 0 "Screen0" 0 0
InputDevice "Mouse0" "CorePointer"
InputDevice "Keyboard0" "CoreKeyboard"
EndSection
Section "InputDevice"
Identifier "Keyboard0"
Driver "kbd"
EndSection
Section "InputDevice"
Identifier "Mouse0"
Driver "mouse"
Option "Protocol" "auto"
Option "Device" "/dev/input/mice"
Option "ZAxisMapping" "4 5 6 7"
EndSection
(unsure both needed it worked)
. 1/5/2025 src14
. /etc/hosts-py.allow (like src11's /etc/hosts-gcc.allow) stops py
inet access anytime used, very fresh not fully tested but can be
out in ./build. build is still not "safe to leave inet plugged in"
however it safeER during build and can be used on installed tbld
(it may not be perfect - may have holes or block unintended sockets
which will be panned out with experience)
. added script to switch between speakers and TV for firefox
. iproute and a few more basic pkg added in prep for future llvm option
. 12/27/2024 sr14 wasn't much effort to make from src11 which is good news
. src11/build ghostscript-8.01 postinstall just after (updated in src11)
cd gnu-gs-fonts-std-6.0 &&
mv -f `find . -maxdepth 1 -mindepth 1 -type f` ..
was supposed to be
mv -f \`find . -maxdepth 1 -mindepth 1 -type f\` ..
. unlike ubuntu RH, Xorg now installs like X11R6 which helps older X apps
run in some cases and i think is easier to manage and work with
. src9-14 gcc-12 opt added to prevent "ET PHONE HOME" /etc/hosts-gcc.allow
. looks like ipv6 fw isn't needed for live or @home use since rules aren't
needed locally but ipv6 rules can be added for special needs as-is
In-depth News for src14:
0 Intel i7 i never did figure out how to boot without Intel IGC EFI VGA
but intel docs assures it has VGA. meanwhile KMS modesetting driver is
said to be "what i7 should be using on all newer intel"
* good news src14: no more problems i had 2017 - 2024 with mm
1 2017-2024 Mathematica cursor glitches or lack of any curor fixed ...
Mathematica 12.3-14 had cursor glitches src9-11 on tbld but also on ubuntu
and win10 2023 nVidia-1050 and i5 HD same. mm14 had NO CURSOR tbld.
(11.3 worked iMac?2017 no glitches any OS, while 12.3 crashed on imac)
No cursor on mm14 tbld ended up being QT_SCALE_FACTOR needed for 40" LED.
Even that removed mm14+src9-12+mesa-13 had cursor glitches plus a
fat short bold cursor randomly in plain text - but somewhere between
10,11,12 on i7 - no more glitches on i7 intel. whish i could say where
exactly the glitch was solved. src11 v. src12 ? i7 v. i5 v. nvidia?
(glitch: extra cursor line left in L margin at times - updating cell
got rid of it, since 13 in all OS, annoying but harmless)
2 mesa-13 has no i7 GL. the Mesa-13 software GL is Terribly Slow with mm14.
(my own 'rayshade-mm' renders allot of the same 3D much faster at better
quality with 'real rayshade raytracer' that should be allot slower than
software mesa EGL mm uses; but never intended as a full replacement)
(infact for i5 personal use I fell back to Mesa-7.8.2 for software GL but
but cannot remember compat with forced xorg upgrades at this time)
THE GOOD:
With src14+mesa-24 on i7 HD (nv-1050 not yet tested on i5 box):
. no cursor glitches at all, indicating WR's "mininum platform" is
not truthful about versions now required (likely Qt and ubu are
is lying to them and they simply don't know)
. definitely hardware accelerated i7 GL, not as fast as nVidia 1050 which
is normal. due to using kernel driver "glxinfo" does not show the
GL vendor string (though all are marked non-slow)
. 1/22 mesa-24.3.0
apparently can build with llvm-only or gcc+llvm - choose either
LLVM-19.1.0-Linux-X64 (official bin), all else (ie spirv) from source.
Mesa-24 always requires llvm to be downloaded and extracted in chrooto3
ie, Mathematica-14 with GL, xpdf, firefox, work nicely.
. 1/11 mesa software GL was NOT working (surprise to me never seen mesa
this f'ed up where software GL didn't work after installed build)
GL Narkive:
I think in junk/ somewhere i mention in 1999 i had glint GL ALL IN SILICON
2M ram worked with mm11 as well as i7 does 2025 (not incl. memory crunch).
Had 0 development time from X11 win95 to get accelerated and or
find an entry point for making a GL program, and no fan! 2025 nv intel
require driver ARE NOT IN SILICON GL requires 25+ extra downloads (some
intently ubuntu only depends won't load .so) and GL is no longer
"in silicon". 9G of llvm to compile a few .so? really can't do gcc?
worse not better (excluding gaming which i don't count as first)
. mesa-24 llvm intel issue (not mesa-13)
The intel modeset may have bug with xorg-damage.
When a window is moved "too far" (on a 40" desktop btw) some
of the window disappears. But if covered uncovered the window is fixed
unless/until "moved too far" again. It is minor: see News above for fix.
Tested as different entirely from the curosr glitches above.
In-depth News for src15:
THE GOOD:
. like src14, the cursor is still clean artifact free
. i7 (youtube) video works even better than in src14 USB
could be wrong seems better visual quality than win10 using nv,
still - full screen is not 60 fps but hey: 4k is allot allot to ask
. i7 Mathematic GL is mostly accelerated, no crashes or glitches seen
which is very welcome (2022 ubu win10 tbld had crashes and glitches
with nVidia 2022 mm13, and iMac had serious crash issues with mm12.3
i had to sell it - while 11.x seemed bullet proof on all and xlfs).
Autodetect doesn't work on i7 or nVidia, never has on Win10 or Linux
"3DRendringEngine"->"OpenGL"
THE BAD: i7 GL Mathematica... (applies to src14 likely also) seems
-> "unaccelerated on certain transparent GL graphics, w/o settings"
-> "turning on hardware depth buffer is accelerated but quality loss"
Really About UDH 770:
. i had certain GL plots in Help slow with nVIDIA at one time which
after one upgrade disappeared. Mathematica, Qt, Xorg can effect.
. UHD 770 apparently has no ACCUM (dropped from OpenGL since 3.1
but nVIDIA 1050 still contains it - "ah, marketing ...")
. complex transparents can't be z-sorted and there will be artifacts
due to this unless a something other than alpha blend is done.
. some articles say complex transparents can be rendered using
gl_accum but anything MODERN should Definitely be using "shaders"
. Mathematica is not always on the forefront of change it may expect
new cards still have ACCUM or be lazy ;) May use EGL 1.3 idk.
. "uncertain" because i am uncertain where the blame is on this one
without bothering WR with a nag question on internals
. very complex graphics lower fps than on nVIDIA, that's normal
. Mathematica's (i7 only) after more testing GL any but very simple
transparent graphics seem not accelerated render slow, complex ones very
slow: 2-5 fps; model of shell growth Applications ParametericPlot3D.
--> Duck Tape Fix:
"3DRendringMethod"->"HardwareDepthBuffer"
But this will cause transparent to have missing or allot of lost depth
resolution (larger triangles instead of smooth surfaces) and on certain
grapics totally not seen objects behind transparents. HDB setting is an
appropriate fix for MolecularModel3D but not appropriate for Mathematic Plots,
but does render complex translucents at high fps. I'm very disappointed I
have no mm provided info to investigate with and with what is it
30 Gig of GL drivers (kernel+built+mm's) I expect better that "EGL software
rendering". These test rob allot of time from consumer fiddling with options
it really isn't fair.
There is GOOD NEWS with transparency is Graphics3D Basic Examples 1,3
renders transparent box better than nVIDIA accelerated (HDB set or not)
because nVIDIA shows glitches i7 does not. (new: could be accum use which
is why the artifacts)
GOOD NEWS with transparency is for Sphere (Bubbles example apply specularity,
opacity, colour, overlapped objects) works fine with HDB set
i7 begins to slow down with >50 objects, unsure if due to xorg memory
allocation (meaning lack of - i gots 16G) or bios video mode or what.
GOOD NEWS with transparency is for GraphicsComplex Neat Examples 100 triangles
translucent surprisingly zippy at huge screen sizes 4k desktop no glitches
with HDB. though slower and looks slightly better with HDB=automatic still
moves with mouse acceptably. Nested cylinders with HDB automatic "seem
different colored, slower" but actually HDB set accelerated is just as good
looking imho - no detail really changed. BoundaryDiscretizeRegion[Ball[]]
has >>100 triangles works high fps at very large screen sizes.
Oddly, 3D Delauny (kinda important) (outer hull transparent with inner shapes)
has more detail with HDB set than "for slow mode" and looks better,
and in both modes there are z-buffer type artifact during mouse, infrequently.
Chemistry molecular graphics sometimes use transparency to show bond type
automatically, they work fast with HBD setting, slow 2-25 fps if not.
Important for me is diffeq graphics (highly complex overlapping random ocean
waves many sharp spikes shallow curves cut off hollow tips to fit in plot).
So far these DO NOT CRASH MM14, which I'm Very pleased with I was about to
toss mm13 nVidia out the window it had crashed often 13 always in 13.x
(soft render mesa-13-24 horrible reverted to mesa-7.8.2 helped allot)
The HDB option is easily flipped between at runtime as a go-between.
EXECUTIVE SUMMARY
Intel Core7 12700K Intel UHD Graphics 770 and Mathematica 14
versus nVIDIA GeForce GTX 1050 gaming Mathematica 13
i7 on tbld src15 had a few dissapointments but were More than offset
by benefits, way more for me.
Have A Text Cursor. mm13.2 on src9-src11 had NO CURSOR w/nVidia
blame anythying i don't care when your customer pays and has no
cursor that's a problem - you can't edit text without a cursor.
Save power, time, money, and PC space and 3dfan issues making a PC
diffeq graphics had crashed often on mm13 nVidia
win10, linux (tbld src9, src11, ubuntu 22)
with src15 i7 and motion "seems cleaner like no wacko glitches during
like nv had" and hasn't crashed yet. that's a waymore factor for me.
No Cursor Glitches that had plagued NV+MM13 and intel mm14 before tbld
updates completed. see src14 news i had extra cursor in margin issue
perhaps newer nVIDIA on mm14 and 550 drv has issues resolved idk i
can't say am_not motivated to "swap card in out" happy as things are
I needed a compiler 8-core thread ripper didn't shop for Xe. At this point
it would not be worth swapping out the UDH 770 isn't much behind and Xe
appears to not support glAccum either.
I should mention nVIDIA drivers are way easier (nvidia setup installs all)
more game beter and don't require allota extra llvm compiling, and *might*
work with mm14, i would say "try within 30 days" incase you have issues with
any new PC build.
AMD had/has a pro series with truly full GL support (all versions not just
one version supported) full color no render fudge errors in silicon
(not gaming junk) for CAD (permedia sigraph) i'm unsure if they still do
they were/are pricey and not best for gaming worth knowing "there is a
difference between GL in silicon and gaming cards".
In-depth News for src16:
. 3/2/2025 src16 (Intel Quick Sink for ffmpeg)
+ src16 does not require uxa tearfree, intel interface fixed
+ ffmpeg more video supported - build opts added
+ Intel Quick Sync Video for ffmpeg (libvpl mfx 2.14)
- src16 still drops frames 4k fullscreen nature youtube 4k desktop,
(needs smaller window size to be naturally pleasant)
- src15,16 lags 720p 4k desktop full winsize (unsure if 240 does)
? i cannot say 100% it's not my wifi lagging cant find free download
. src16 (if not 15) does QHD 1440p "ok" on a 4k
desktop using matching win size of 2560 x 1440 @ 30fps or 60fps.
(less frequent lags). smaller win sizes have no lags.
+ it can do @60 fps "but not perfect" larger than 2k base window size
no matter the requested 4k 2k youtube setting, not real 4k but nice
+ 1440p can be done smooth fullscreen like this: abandon 4k desktop.
xrandr --output HDMI-1 --mode 1680x1059 --scale 1.5 (roughly 1.5)
(see /usr/X11/var/log/Xorg.0.log for available modes)
LED settings -> Full (works if your TV supports speed scaling)
now 1440p full size full screen no lag 2500x1500 30fps
but 60fs still wants 2048x1080 to be no lags, 60hz led btw
. GL mathematica no noticeable difference
+ NV's GAINS smooth full screen speed < 1400p in src16 Versus src11,
(minor differences in still frame quality as to i7). but src9 had this
with < ff99 lost in ff99-ff130, now is back in src16. hmmm. a + ?
+ nv-installer easy as usual (requires linux build dir must build src16)
GL is uneffected depends on official 550.107.02 driver pack
. NV's youtube performance 1910x1051 on 100hz led is NOT better than
Intel UHD 770 on 30fps or60fps nature youtube (yr 2017 v. 2021 btw)
(needs small win size reduction to be smooth). Intel has been seen
smoother at times, but isn't always.
. 3/17/2025 slack news is that Slackware 15 inird.img was copied out to
a directory, chrooted, pkgtool or installpkg installed to partition, and
simple things like bash, netwatch, xpaint work inside chroot or outside
in host OS (src16's). todo: no simple script to do that, though anyone
familiar with these things would have no problem no tbld workarounds
were needed. HOWEVER - it's not a convenient way to do things especially
for newbies - this moves me to add more optional downloads (todo
and perhaps a download list for that)
. easy to pkgtool pkgs into chroot - but X apps wont run in chroot
if tbld's X is already running, and increasingly the only thing
you need ubuntu slack15 for is the megawares requiring X
(thinking future - ie, slack Xpaint is xorg-only stuff and they
are removing old text based goodies not adding these days)
though - there's still allot slack can do in a chroot
for advanced users - a good addition as to chroot additions
. not easy to copy any support files from slack chroot to tbld,
too much bother for most, by typical as to doing so and works
a few short scripts to kick this off since it's too complicated to
package as to where things stand currently:
#!/bin/sh
set -e
pwd="$PWD"
[ -z "$1" ] && {
cat << EOF
$0 target_dir [src_dir]
target must be a directory that can be created
source/isolinux/initrd.img ; dir slack usb is mounted on
source must be mounted, but will be dis-mounted so setup can choose usb
EOF
exit
}
[ -n "$1" ] ; target="$1" ; [ -d "$target" ] && rmdir "$target" ; [ ! -d "$1" ]
source="x" ; [ -n "$2" ] && source="$2" ;
[ ! -f initrd.img ] && {
[ -f $source/isolinux/initrd.img ]
cat $source/isolinux/initrd.img | xz -d > initrd.img
}
[ -d "$target" ] && rmdir "$target"
mkdir "$target"
cd "$target"
cpio -idm < ${pwd}/initrd.img
umount "$source"
## xxx do no "run: chroot $target /usr/lib/setup/setup"
# did this after extracting initrd.img to y/
#----------------
chroot y
# do some /dev copying and fstab hacking borrowing from tbld
mount -a
mount /sys
mount /dev/sd(slackware-usb) z
mount /dev/sd(target gpt part) mnt
cd z/slackware64/a
/sbin/pkgtool
cd /
chroot mnt ldconfig
chroot mnt
#ok
#----------------
menu:
[x] from current directory
the tool begins asking each pkg in order and installs them to /mnt
it worked nicely! (as far as i am concerned) ran netwatch for example.
from a different xterm in host OS i made a couple of requsted softlink from
mnt/ in /usr/local and the host OS could also run netwatch without a problem.
x/oclock and x/xpaint work out of the box using hostOS (haven't tried
installing X in slack since tbld uses it's own)
Up and coming pains in tbld:
. WiFi "list access points" iwl displays spam message: "will not work on
Intel WiFi7". If that's true you must know the simple string to type in
since it won't be displayed.
. I am interested if TBLD does not work in "most apartment complexes" due to
WiFi due to typical access or if something major on laptops does not
work other than powerd (you'll have to add powerd i have no laptop to
test what works on any given top). Leave note on sourceforge if you please.
. Microsoft's GITHUB refuses GIT access without newest ubuntu desktop running:
I know don't tell me that's why i don't like big tech war on linux.
Extra Questions not in main README.txt
Question: do i need USB to boot
. YES - explained more in README.build-0.2
. first boot USB option boot to usb
see /README and follow instructions to install
. kernel, modules, and boot on USB must match those on HD to boot to HD
which after hd-hd-install, they will
. after install select boot to sda4?
if you didn't build USB yourself you'll have to edit that to sda1,2,3
in grub or EFI editor the first time
you could also edit it on using vi before booting
vi /boot/grub/grub.cfg # edit the one on usb
mount the the USB's EFI somewhere
vi /EFI/boot-hd.nsh
. tbld boots kernel from USB always even if booting to HD or USB
this is good it means TBLD will never accidently change your boot
in a way that wrecks your win10 booting and that ubuntu cannot
"write over your boots" `by mistake' since always your USB will boot.
it also avoid that "raid partition not working" issue - tbld is not
equipped to hyper analyze your custom disk structures and doesn't want
to be in that loop
Question about grub or EFI kernel parameters and initrd:
. dont remove rootdelay=5, edit it to 7 maybe
. linux needs a few to several seconds for bios info calls to finish
but it doesn't wait unless asked to
. you will see "VFS: Cannot open root device unkown-block(0,0) PANIC
if your delay isn't long enough
. tbld sets correct block uuid of USB and puts it in GRUB and in foo.nsh
. non-intel motherboard may need initrd option for non-intel ...
xhci-hcd.ko ehci-hcd.ko uhci-hcd.ko ohci-hcd.ko ../storage/usb-storage
(non-intel hcd in past kernels can jam intel hcd, also initrd is not
always a reliable thing between various kernels, current kernels
do it ok though)
Question: i hate english it uses --no-nls and doesn't build uber moxydox
;) not my fault !
. building from scratch means unreliable (often malicious) tools cannot
be done because they stop build. tbld starts with as little as possible
to insure it can be done (still) and to insure against side-band depends
. UTF8 tools ad a ton of bloat (10x the size, 1/2 compiling speed)
. it'd be nice if at END of build to go back run "make doc" on each
i looked to do it and found it'so hacky, not-standardized, no way to
do it except trivially "on many with -i -k" - which can end up harmful
btw and even end up replacing os compiler with broken one
. really better to make what you use/need. just rebuild pkgs you need as
--disable-nsl-doc (see build.global.opt)
. at end of build is ok to add optional moxydox apps, though tbld has no
optional section for that currently
Question: it builds some un-necessary old apps near end of build
. yes but it tests they still work with glibc changes and if you
don't build them you still have the recipes that show tricks
you'll need if you need to use older software sometime
Question: how to update kernel
HOW: you need kernel headers before building an OS for a new kernel
and you need that kernel booted before building
HOW: build of 6.15.5 was done in src16, tarball of headers taken, boot
files copied to usb16, booted in X w/o affair on intel UHD (old firmware
and kmod-32 worked, sometimes these must be upgraded before reboot), then
src17 was compiled. core2 grub boot was made also, not yet boot tested.
NOTE: you cannot do this if moving from 32 to 64 you need to begin by
booting (a, any) 64 kernel then doing HOW
Reference of what main libs are in which srcN
src11:
. gcc-12, glibc-2.38. binutils-2.42, linux-6.10.2, xorg-server-21.1.4
. see Errata about src10-14 being "steps toward src16"
see sources-all.txt for all newer versions for things like gtk
src14: pre-release - slightly upgraded src11 - use src15
need download:
sources-src14-additional.tar
tbld-src14-2025.tar.gz
optional download (to build mesa-24 intel req'd):
LLVM-19.1.0-Linux-X64.tar.gz (official) copy to /mnt-src/usr/src/dl
in list.test.3 uncomment line ## need_llvm--__foo
(can be done after build finishes or before started)
gcc-14.2 glibc-2.40 crypt-2.38 and some build chain; automake flex math
mesa-13 or mesa-24.3.0 with either software GL support or intel(s)
built with llvm and all else from source, optionals
tbld's kernel intel driver is hardware accelerated in Mathematica 14
same xorg-21 apps and kernel as src11
--> see News below and necessary xorg.conf entry
src15: pre-pre-release - slightly upgraded src14
gtk-4.16.7 gtk+-3.94.0 + several bumps needed to do that, vulkan.so
list.test.3 probably needs more optimal re-arranging but builds,
gtk4 now last built lib to soak up gl a/v more depends
--> see News below and necessary xorg.conf entry (intel UHD 770 not nv)
src16: pre-pre-pre-release - slightly upgraded src15
Extra Need Tar download is currently In Addition to src14:
sources-src16-additional.tar
tbld-src16-2025.tar.gz
xorg-server-21.1.16 and x.org/individual all updated similarly
built 1x only, not run-tested yet (xorg Not tested to startup yet)
--> see Errata about src10-14 being "steps toward src16"
src10: lesser versions than src11, same kernel.
gcc-12, glibc-2.34, binutils-2.32
src12: lesser versions than src10, same kernel.
gcc-10, glibc-2.31, binutils-2.52
src20: see below, re-release of src2 but has new adds (EFI, live usb)
src21: builds Xfree86-4.3.0 not Xorg, lesser mesa-ver, otherwise same
src9-2024
. Xorg-1.17.4, 32+64 gcc-10.2 glibc-2.25, meson built (many libs recent
or latest), linux-5.11.8
. Mathematica 13.1 runs faster, firefox-109 has Youtube video audio
(src11 runs mm14)
. same as src9-ff109 but with EFI and other new stuff
src8-ff99
. HAS BEEN PUT ON HOLD (intermediate between src2 src9)
. ff99 binary video+audio, mathematica-13.1, Xorg > 7.7, 32+64
work (intelHD nvidia PC), is notably "the last non-meson build"
gcc-4.7 and gcc-10.2, glibc-2.25, mesa-12, linux-5.11.8
. this did NOT run Mathematica 13 that had been it's goal
src2-ff44 now named "src20" (xorg) and "src21" (Xfree86-4.3.0)
. ff44 and qt5.1 build from source, 32+64, no audio video
(maybe can be done just snipping from src8's list but ff44 may ignore, idk)
X11R7.7 exactly, gcc-4.7 glibc-2.11 linux-4.16.12
. efi kernel config not available as list item
----------------------
really extended questions - junk questions
Question: GAMING
. sourceforge building
. java
. beware games like RPGBoss.java have class but little .java
they are not open source many .class few .java, have onerous
licenses hard to find, dom->apache-xerces2->junit->ant->grettle
hundreds of .class no documentation no source. you cannot compile
and re-distribute such "java games". not open.
. these warez always build fail - secrets are many
. they also are deleting apache2 java libs source code
so getting them imppossible
. if you see sax or grettle run the other way
if javac *.java (or build.sh) doesn't work: its not java
. biege worked after a little classname hacking, but badly designed
not worth it - worked same as .jar which was bad
. no RPG java i could find - funny i thought there was some in 2000's
. old bug good SDL games - SDL-1.2 era
beware debian SDL (ie freedroid) requires SDL_image but SDL2 which
"obviously you want installed" conflicts and prevents build, typical
malware to "remove assets for people into unity win10 rh suse et al"
also they've been deleting - SDL1_image was removed/block on github
freedroid had been a classic for RH is now asset deleted
Question: why no gnome?
. older xlfs woody gnome list worked, was unwieldy to build, very
anti-climatic w/o all apps of same era: gnome-print wont work today's
printers (need gnome4), gnome-panel mostly does nothing by itself ...
Motif is nice, has file-box and stuffs if sought, so no point.
. KDE qt is very large, is a password stealing kit, is just X wrappers,
"ransomware abandonware group outside usa seeking strong ARM sales"
if my arm isn't twisted (if mm ff xterm java runs) i aint gonna bother
Question: why dont u use .git ?
. git builds fail ALLOT - you have to run same extact ubu flavor as
developers (who've often hosed their machine so it's impossible)
. src8 xlfs-0.1 tried that
it wasn't months until i had to provide alternate URL for certain
very high profile sources and in a year i saw about 20% of all URL
were no longer valid
. in short: it was a waste of my time - had i just released the sources
i used (i try to prevent upgrade in places) and only that for major
releases - i saved time
. assure others could use it and assure against big tech microsoft assaults
ie - grub-0.95 i'd used early before xlfs-0.1 became deleted from
http://savannah.gnu.org/
(actually still curls http that's good programmer friendly btw)
. .git and other can not only delete past versions (and have) to make
"other os difficult" (they delete the best ones leave ones with awful
problems). git source are huge compared to source.tar.gz - bloat.
they can use "security edits" to claim it had to be hacked to
"protect you". that means build fail that means what TBLD don't like
(you know the mob and protection money)
. sourceforge can't store whole git histories of every project multiple
times for each linux OS - i'm lucky they aren't senting me email about
my current uploads
. good code has CFLAGS that enable disable features nicely and portably
so you DON'T NEED an rcs on you have all well designed working version
of old and new version all in one source. and if that don't work, then
already there are multiple versions and you likely only want the one
you actually use. (storage silo can keep the rest - meh)
. compained of in README.build-0.2, microsoft prevents use of git without
git upgrade that requires depends on GUI junk from ubu rh and deb
(they've made git access a virus)
Future thoughts so far:
tbld is a base OS builder not aimed at abandon-ware collections.
TBLD does not want "to be yet another OS authors must make packages for
that only work one release".
TBLD can't necessarily build "gigawarez designed platform only" (Konqueror?)
which increasingly ebb toward non-unix structured and may not run apps
unless systemd booted, and they tangle crypto-ver into (many things) where
it's impossible to have the same combo for more than one plat one release
(ABI should never change - but they have broken those rules now also)
(domination ware, war on linux, and so on) - these are designed domination
malwares which copyprotect by secrets and mutations time bombs.
TBLD wants to work potentially with musl libc (perhaps they'll have a
binary ABI guarantee Sun used to have), or even work with Sun. TBLD wants
history *supported* but does not want OS building decided or stuck on
particular "BIG TECH bribed agenda teams". Since TBLD started 2010 there
have been other OS builder startups probably for similar thoughts.
(musl currently cannot work with most of tbld - but may someday)
src16 will run some bins out of box of slackware ubuntu, but others
require special libs or ie /usr/share/foo files - so it is not convenient
it requires chrooting and copying files between those
TBLD is @home dev which locally (to me) means Intel which means I can
run what I release - I have no lab or team to test ARM and there has
a ton of focus on ARM in linux (they don't need help) and to such a point
they destroy Intel builds while inserting ARM script in Intel's build
sections (ie, grub) while side-stepping same (ie no OBJTOOL for ARM in lk).
Similarly, I have no AMD stuff at this time.
TBLD has no ipv6 cloud, it is not a pre-configured for secure cloud host.
Steam support: if Slackware kept it out and i put it in, it was by accident
I like portability unix std that last forever. I'd still be using XFree86
(with optional modules) if it weren't for HARDWARE DRIVERS and how those
are used to "delete standards" for malicous purposes. Can't run without
hardware - not @home, no.
Download Latest Version
src17-32G-gpt-2025.usb.gz (2.7 GB)
