[gentoo-user] Cannot emerge 32 bits version of gnome libraries
Hi, I have build failures for various gnome libraries. The errors happen when compiling the 32 bits versions, and seems to be due to "-m32" not being used for the C compiler. The errors are either the linker complaining about architecture mismatch, or glib static assertion errors about pointer sizes. I can see from the build log that the "-m32" option is missing. In particular, in the prepare phase, I see the following: C compiler for the host machine: x86_64-pc-linux-gnu-gcc (gcc 9.3.0 "x86_64-pc-linux-gnu-gcc (Gentoo 9.3.0-r1 p3) 9.3.0") C linker for the host machine: x86_64-pc-linux-gnu-gcc ld.bfd 2.34 C++ compiler for the host machine: x86_64-pc-linux-gnu-g++ -m32 (gcc 9.3.0 "x86_64-pc-linux-gnu-g++ (Gentoo 9.3.0-r1 p3) 9.3.0") C++ linker for the host machine: x86_64-pc-linux-gnu-g++ -m32 ld.bfd 2.34 So strangely it's present for C++ but not for C. Also, I can emerge other non-gnome multilib packages just fine. Has anyone encountered this before? Any tips for debugging that? Thanks, François-Xavier
Re: [gentoo-user] Upgrade an old system
On Mon, Dec 14, 2020 at 02:57:22PM -0700, the...@sys-concept.com wrote: > [snip] > > emerge --update --oneshot portage > > !!! All ebuilds that could satisfy > ">=app-crypt/openpgp-keys-gentoo-release-20180706" have been masked. > !!! One of the following masked packages is required to complete your request: > - app-crypt/openpgp-keys-gentoo-release-20200704::gentoo (masked by: EAPI 7) > > The current version of portage supports EAPI '6'. You must upgrade to a > newer version of portage before EAPI masked packages can be installed. > (dependency required by > "sys-apps/portage-2.3.99-r2::gentoo[rsync-verify,-build]" [ebuild]) > (dependency required by "portage" [argument]) > > I don't even have "app-crypt/openpgp-keys-gentoo-release" installed why is it > complaining? > I've done a similar update recently, it's definitely possible if you know what you're doing. A few things: First, you need to update portage. To do so, you'll need to manually downgrade the portage ebuild from EAPI 7 to EAPI 6 (simply change the number in the ebuild). First, emerge it without rsync-verify. Once you have portage up to date, it'll support EAPI 7 and life'll be easier. You can then reemerge it with rsync-verify, as the EAPI 7 dependencies will be installable this time. Then, I'd recommend upgrading the toolchain (binutils, gcc, glibc), otherwise you might run into compilation issues. After that, you can update packages. For big updates like this, I like to do it by chunks as it's easier (and it makes reading the emerge plan less painful). Like all the perl packages, then all the python packages, and so on. For that, the "emerge by path" feature is convenient; for example to update python packages: emerge -au /usr/lib/python3.7 In case force-unmerging python modules makes the update easier (resolving blocks), you can do so. That's actually safe to do, portage will still be able to emerge packages. Python itself (the interpreter) is the only hard dependency of portage. Hope this helps, and good luck with your update!
Re: [gentoo-user] Cannot start xfce from lightdm [solved]
On Sun, Nov 01, 2020 at 04:18:23PM +0100, François-Xavier Carton wrote: > Ah, this might be the problem. I don't have that file, I only have > 00-xhost in there. I checked the dbus ebuild, I need to enable the X use > flag. Will do, and report back :) Thanks for the pointer! > Yes, that solved the problem. Thanks a lot!
[gentoo-user] Cannot start xfce from lightdm
Hi, I'm installing gentoo on the family computer, and I'm opting for lightdm and xfce (for ease of use for non-experienced users). As per the wiki, I emerged the packages, enabled and started the dbus and xdm services. Lightdm starts fine, but xfce fails to start with the error: "Unable to contact settings server; Unable to connect: connection refused". A few things I tried: - starting another DE (dwm) from lightdm: works - starting xfce from startx: works - starting xfce from a running x server (via terminal emulator): works - remerging xorg with "elogind -suid": still the same (for simplicity I'd prefer "suid -elogind" though) What I noticed: - the error seems to be dbus-related (from the source code, it seems xfconf_init fails, which in turns calls dbus things); however I double checked that dbus is enabled and started. - when xfce starts successfully (eg. via startx), it starts xfce4-session with dbus-launch (not sure where that happens; I see no mentions of it in startxfce4 nor /etc/xdg/xcfe4/xinitrc), whereas when started from lightdm, xfce4-session is started normally (without dbus-launch) I'm not used to graphical login managers, and I'm running out of ideas. Does anyone have some idea of what I could try to get xfce running from lightdm? Thanks, -François-Xavier
Re: [gentoo-user] Cannot start xfce from lightdm
On Sun, Nov 01, 2020 at 02:39:02PM +0100, netfab wrote: > Le 01/11/20 ?? 13:32, Fran??ois-Xavier Carton a tapot?? : > > I'm not used to graphical login managers, and I'm running out of > > ideas. Does anyone have some idea of what I could try to get xfce > > running from lightdm? > > It works fine here ?? (xfce + lightdm). > > Do you have : > > session-wrapper=/etc/lightdm/Xsession > > Enabled in /etc/lightdm/lightdm.conf ? > Yes. I did not edit this, so I think it's the default config. > /etc/lightdm/Xsession is sourcing > /etc/X11/xinit/xinitrc.d/80-dbus > > which runs dbus-launch. > Ah, this might be the problem. I don't have that file, I only have 00-xhost in there. I checked the dbus ebuild, I need to enable the X use flag. Will do, and report back :) Thanks for the pointer!
Re: [gentoo-user] Re: Building packages in different prefix without rebuilding system packages
On Fri, May 15, 2020 at 12:53:16PM +0200, Michael Haubenwallner wrote: > Hi François-Xavier, > > What you're after is known as "Prefix/Stack", where you have some "base" > Prefix, which's portage does manage packages in another - stacked - Prefix. > > While this does work already with "Prefix/Guest" as base Prefix, there is no > technical reason to not work with "Prefix/Standalone" or even "Gentoo Linux" > as the base Prefix. The only problem maybe is that "Prefix/Guest" does use > a portage version with additional patches. > > But still, you can get an idea about how this works, using "Prefix/Guest": > > $ docker run -t -i gentooprefix/prefix-guest-64bit-fedora -c /bin/bash > > At the docker prompt: > > Enter the Guest Prefix preinstalled to /tmp/gentoo: > $ /tmp/gentoo/startprefix > > Set up your stacked Prefix: > $ prefix-stack-setup --eprefix=$HOME/stack > --profile=/tmp/gentoo/etc/portage/make.profile > Note that "~/stack" will not work here, bash refuses to resolve '~' after '='. > > Leave the Guest Prefix: > $ exit > > Enter your just created stacked Prefix: > $ ~/stack/startprefix > > Emerge your package, for example: > $ emerge sys-libs/zlib > > Have fun! > > HTH, > /haubi/ > Thanks, this looks great! I'll play with it and see how it works :)
Re: [gentoo-user] Building packages in different prefix without rebuilding system packages
On Thu, May 14, 2020 at 09:26:10AM -0400, Michael Orlitzky wrote: > On 5/14/20 7:55 AM, Neil Bothwick wrote: > > On Thu, 14 May 2020 18:17:06 +0800, Pengcheng Xu wrote: > > > >> That seems interesting. Do we need to include Portage install prefix > >> (/var/tmp/portage/category/package/..., the image path prefix before > >> actually merging with /)? > >> > >> Regards, > > > > No, just the --prefix=/home/blah/ that you want added to the ./configure > > invocation. > > > > This is a good way to install packages that you've built by hand into > (say) your home directory, but it will cause problems if you try to > trick portage into doing it. The big problem is that no other packages > are going to know where to find the thing you just installed. Everything > else in the Gentoo repository is designed to use standard values of > PATH, LD_LIBRARY_PATH, the compiler's include dir, PKG_CONFIG_PATH, etc. > If you take one program and put it somewhere non-standard, then every > package depending on it is going to break. > > If you install an *additional* copy (built by hand) in your home > directory, that's fine -- the system copy will still be in the right > place -- you just don't want to hide the system copy where nobody can > find it. > In my case, this wouldn't be a problem: I don't want the packages to be accessed by anyone, just one user. I can set PATH, LD_LIBRARY_PATH, PKG_CONFIG_PATH and MANPATH for that user. I already do that for things I build manually anyway. EXTRA_ECONF is nice, I didn't know about it. It looks like MYCMAKEARGS can be used for cmake ebuilds. For other build systems, it might be necessary to edit the ebuild, or set different variables. I still kinda think that being able to install with a prefix (like EPREFIX) but using the base system would be a nice feature. As discussed previously, there would be the problem of updates; but it isn't very different from installing software manually. If I clone something and build it locally, a world update might break break it, and portage cannot rebuild it automatically since it's not aware of it. This is why I think it would be nice if portage supported it; that way, after an update of the base system, updating the prefix system would solve the problems. It's probably difficult to implement that in portage, though.
[gentoo-user] Building packages in different prefix without rebuilding system packages
Hi, Is there a way of installing packages in a different prefix while still using system packages? I've tried setting EPREFIX, however doing that will install all dependencies in the prefix, even if there are already installed in the system. I was hoping to install some packages in user directories, but I also don't want to duplicate the packages installed globally. For example, most packages eventually depend on gcc, which I definitely don't want to compile twice. So ideally, only dependencies that are not installed globally should be pulled in. I was not able to find a way of doing that, but I feel like it shouldn't be too hard, because EPREFIX almost does what I want. Does someone know if it's possible without too much tweaking? Thanks, -François-Xavier
Re: [gentoo-user] Building packages in different prefix without rebuilding system packages
On Thu, May 14, 2020 at 09:07:43AM +0100, Michael wrote: > On Thursday, 14 May 2020 06:13:33 BST Dale wrote: > > François-Xavier Carton wrote: > > > Hi, > > > > > > Is there a way of installing packages in a different prefix while still > > > using system packages? I've tried setting EPREFIX, however doing that > > > will install all dependencies in the prefix, even if there are already > > > installed in the system. > > > > > > I was hoping to install some packages in user directories, but I also > > > don't want to duplicate the packages installed globally. For example, > > > most packages eventually depend on gcc, which I definitely don't want to > > > compile twice. So ideally, only dependencies that are not installed > > > globally should be pulled in. > > > > > > I was not able to find a way of doing that, but I feel like it shouldn't > > > be too hard, because EPREFIX almost does what I want. Does someone know > > > if it's possible without too much tweaking? > > > > > > Thanks, > > > -François-Xavier > > > > I'm clueless on EPREFIX but if you want to avoid compiling a package > > twice, would the -k option help? If you have portage set to save the > > binaries you compiled before, it would install from that instead of > > compiling things twice. > > > > Just thought I'd mention just in case it would help. > > > > Dale > > > > :-) :-) > > The whole concept of EPREFIX is predicated on installing a Gentoo system > within a different path/filesystem than the / of the host installation and > being able to run it as a non-root user. As I understand it, using libraries > from the main system and potentially altering them in the process is not > going > to work without changing Gentoo's eprefix intended design. > > It should be possible to change the prefix paths selectively, in particular > the LD_LIBRARY_PATH to link binaries from within the prefix to libraries in > the host system, but I'm not sure what privileges are needed to install/run > such a hybrid linkage and how an update of the host system will break > installed packages within the EPREFIX and vice versa. We're talking of a > Frankenstein build here with the potential of install operations on one > system > would be breaking the other, including portage itself. > > With containerisation of applications there may be easier ways to achieve > what > François-Xavier is looking for. I am thinking of running sandboxed > applications in the likes of flatpack, snap, zero-install, appimage and > whatever else may have been devised lately. However, with these systems you > end up using what's already been developed and any static libraries their > devs > considered desirable. If you want a bespoke installation optimised for your > hardware and chosen compilation flags, then you are probably looking to > develop a containerised application for your own use. > > Someone more knowledgeable in both Gentoo's EPREFIX project and containerised > apps should chime in soon to offer more helpful advice. I wouldn't try to install different versions of the same package, just add new packages to the base system. But I can see your point about updates; since the base system is not aware of the prefixed one, it cannot rebuild the packages in the prefix automatically when needed. It's indeed more complex than what I hoped :(
Re: [gentoo-user] 32GB RAM and Swap
On Sat, May 02, 2020 at 10:12:08PM +0800, William Kenworthy wrote: > I am afraid this is an ".. it depends" question. > Yes, I agree. > If you work with large images or data sets, swap can be really handy. > If you are doing a little programming, web browsing, reading email you > will *probably* be ok, but why risk it? > Risk what? Having the OOM killer kill the problematic process? Depending on your usage, this might be the best. Personally I prefer that to a system that is stuck. I never had to force reboot on a system without swap, whereas with swap I had to reboot most of the times swap was used. Also it's super annoying when your system freezes because of a background process swapping (eg. an emerge world update) while your doing something else. I've been running a 8GB system for a year, before that a 4GB system. Both without swap. It's been fine so far. I did have processes killed (eg. firefox compiling), but at least I can continue to use my system without being interrupted by a freeze. It's a lot less frustrating to have to resume a killed compilation than to deal with a frozen system. > I have a 32gb ram in a master server for an mfs filesystem - it normally > sits at about 5GB of ram - however it can go well over 32Gb into swap at > times - the first machine I tried it with only had 4gb ram and crashed > when it filled the ram, and 8g swap taking the test file system with it > - its now production so I am not going to risk it by underprovisioning > swap. My 32Gb desktop is not using any swap at the moment ... but it > has used it at times. > > So, yes its quite likely you wont use swap - but if you do something > that needs it, it can help avoid a very messy crash. > > Swap is slow, but if you actually need it - its probably critical that > you have it! Unless you are really short of disk space, treat it as > insurance :) > > Look into using swapfiles instead of partitions for flexibility, and the > sysctl values of "vm.swappiness" and "vm.vfs_cache_pressure" to manage > swap usage (you can set to not use swap until it really has to - some > have seen the kernel being too eager to swap out causing slowdowns, > though you can make it go in the other direction and "thrash" when it > actually needs to use swap if you go to far. The default kernel swap > mechanism isn't really that bad! > Swapfiles are great, because you can only add them when you need and remove them when you're done. I sometimes use them when emerging large stuff when I have other big processes in memory. > So yes, most of my machines don't need swap *right now* and swap looks > like its not being used so it could be removed, but I cant guarantee > that they never will, and having years of experience using swap I > recommend that its better to be cautious and survive :) My systems usually survive (are not forcibly rebooted) better when there is no swap. But I agree that in the end, it depends on the usage.
Re: [gentoo-user] g-cpan equivalent for python
On Sat, Nov 23, 2019 at 11:21:46AM +0100, Dan Johansson wrote: > > OK, thanks for the feedback. > I will look into creating a ebuild for this (not my first ebuild, but > the first that has anything to do with python). > But, I will first try to build it in a "protected" environment like > Ralph suggested just to see if it works at all (would not be the first > time some software does not work is I would like it to). > > Regards, > -- > Dan Johansson > *** > This message is printed on 100% recycled electrons! > *** > You can also use "pip install --user package", so that the files are installed in ~/.local and don't conflict with the system ones.
Re: [gentoo-user] vimrc in (g)vim and highlighting
On 10/3/19 10:23 AM, Samuraiii wrote: Hi list, I have strange problem with my ~/.vimrc file. When I have it (does not matter if it is empty or not), the syntax highlighting is not working at all. But when I delete it, vim highlights as expected (I have already tried to set "incompatible" mode in vimrc with no success). Cheers S Do you have "syntax on" on your .vimrc? If you do and it's not working, you can debug what is executed at startup by running "vim -D", and check that no plugin set "syntax off" or "set compatible". Hope this helps.
Re: [gentoo-user] escape from i3lock
On 7/10/19 7:03 PM, Ian Zimmerman wrote: Here is my next "low information" question, haha. I use i3lock which is like Xscreensaver but much much simpler; it plays no movies or games, just blanks the screen with a configured color or image. To unlock it you have to type your password. It bothers me that even when i3lock has locked the X session, I can still switch to other Linux virtual consoles with Alt-Control-F , without typing the password. It so happens that on one of the other virtual consoles there is often an interactive root shell :-P So, is it possible to prevent virtual console switching while the X screen is locked, but still allow it at other times? Looks like something the locker program would have to do, not the X server; but again I don't know much about this stuff. Not a direct answer to your question, but as a workaround you can use tmux sessions, and simply detach them and logout when you lock your computer. Also, if this is just a shell to start the X server, you can launch it as "startx & bg; disown" and then logout.
Re: [gentoo-user] UEFI kernel installation?
On 6/17/19 5:37 AM, Grant Taylor wrote: I doubt it. I've routinely done emerges on machines with < 16 GB of memory and 2 GB of swap.?? Including llvm, clang, gcc, rust, Firefox and Thunderbird. I routinely do an emerge -DuNe @world on a VPS with 1 GB of memory and 1 GB of swap.?? It works just fine.?? If I want to speed things up I enlarge the VPS to 2 GB of memory and 1 GB of swap.?? Granted, it doesn't try to compile things like Firefox and Thunderbird, thus Rust. Yes. I've run systems without swap and <=8GB of RAM for years and haven't had any problems with emerging packages (including rust, firefox, gcc...). If hibernate is not needed, and the system has enough RAM, there is little point in having swap at all. It all depends on the usage, so I don't think any (generic) guidelines make sense.
Re: [gentoo-user] OT scripting - strip zero if between period and digit
Le 23/01/2019 à 04:19, Adam Carter a écrit : > François-Xavier > > My bad, it should be: sed 's/0*\([0-9][0-9]*\)/\1/g' (tests are indeed needed!) Many thanks François. This is almost right, but it is also stripping zeros that follow a letter, and I only want it to strip zeros that are proceeded by a period. There are no leading zeros in the first octet of the IP so that case does not need to be handled. Does the \1 refer to what's in the ()'s? So anything that one would wont to carry through should be inside the ()'s and anything that's outside is stripped, right? Yes, \1 is the content in (). But adding letters inside won't solve the problem, eg. "a01" will still be changed to "a1". AFAIK, there is no way to express "start of line or a character" in sed, but you could do two regexps, one starting with ^ (start of line), the other with \. (dot) sed 's/^0*\([0-9][0-9]*\)/\1/g;s/\.0*\([0-9][0-9]*\)/.\1/g'
Re: [gentoo-user] OT scripting - strip zero if between period and digit
Le 22/01/2019 à 03:05, François-Xavier CARTON a écrit : Le 22/01/2019 à 00:50, Adam Carter a écrit : I need to clean up a file which has IP addresses with leading zeros in some of the octets so I need to make, say, .09 into .9 How do i do that in sed/awk/whatever? I believe that should do: sed 's/0*\([0-9]\)/\1/g' eg. $ sed 's/0*\([0-9]\)/\1/g' < My bad, it should be: sed 's/0*\([0-9][0-9]*\)/\1/g' (tests are indeed needed!) François-Xavier
Re: [gentoo-user] OT scripting - strip zero if between period and digit
Le 22/01/2019 à 00:50, Adam Carter a écrit : I need to clean up a file which has IP addresses with leading zeros in some of the octets so I need to make, say, .09 into .9 How do i do that in sed/awk/whatever? I believe that should do: sed 's/0*\([0-9]\)/\1/g' eg. $ sed 's/0*\([0-9]\)/\1/g' <
Re: [gentoo-user] Xorg on really old PC
Le 31/08/2018 à 05:10, Walter Dnes a écrit : Given the inter-connectedness of linux libraries, you're not going to get an ancient version of mesa to work with a new Gentoo install. That's where Gentoo helps, being a source-based distribution :) Forget about OpenGL and undo the masks. Another option might be to set... VIDEO_CARDS="fbdev" ..in make.conf, and use the xf86-video-fbdev driver and see if that works. Thanks for the suggestions. I'll try one more time with xf86-video-intel first (using the same package versions as debian squeeze for the xorg-related packages), as I would like to have OpenGL. But I'll try unmasking and fbdev if that doesn't work. Le 31/08/2018 à 12:01, (Nuno Silva) a écrit : > Perhaps try an older linux kernel, from the same time as the versions of > X11 and mesa that you are trying to run. > > Running a system like this is going to be harder now, though, because > the X11 headers were reorganized, and packages which use X11 now depend > on the newer headers, which will most likely pull in the new > xorg-server. I'll try kernel 3.2 and xorg server 1.7.7 which are the versions used in debian squeeze. This is the last debian release with mesa < 8.0.
[gentoo-user] Xorg on really old PC
Hi, I'm installing Gentoo on a really old PC (with a Pentium III and a i815 graphic card). I've installed a minimal Gentoo system, and I'm trying to get Xorg working. I've searched on the Gentoo wiki, and found that I should use an old version of mesa (<8.0) to get OpenGL support [1]. So I have masked recent versions of mesa and some Xorg packages in order to install the last version of mesa that is below 8.0. My package.mask is: >media-libs/mesa-7.10.3 >=app-eselect/eselect-opengl-1.2.6 >x11-proto/glproto-1.4.15 >x11-base/xorg-server-1.12.4-r5 I have no xorg.conf. When I run startx, the screen goes black and the Xorg server segfaults, according to the log file [2]. Nothing works after that, including the keyboard, so I cannot go back to a linux console. Maybe it is a configuration issue, or maybe I'm using a buggy version. Has anyone any suggestion on what configuration and versions I should use? Thanks, François-Xavier Carton [1] https://wiki.gentoo.org/index.php?title=Intel=599748=599742 [2] Xorg.0.log: http://sprunge.us/ZdWJNH
Re: [gentoo-user] Heads up: Gentoo fouls up mail transport agent.
Le 22/07/2018 à 00:10, Mike Gilbert a écrit : On Sat, Jul 21, 2018 at 5:03 PM, Alan Mackenzie wrote: Hello, Gentoo. Right at the moment, I feel a lot of sympathy with Alan Grimes, and need a lot of restraint in avoiding the use of swear words in describing some Gentoo developer. ... nullmailer installs a file /usr/sbin/sendmail. This masks out the correct /usr/bin/sendmail (which is a symbolic link to s/qmail, which I installed by hand, not using emerge) because /usr/sbin is before /usr/bin in $PATH. ... But what's the proper method to tell my gentoo system that I don't want crud like nullmailer installed? How can I guard myself against such presumptiousness on the part of the Gentoo devs in the future? You must have installed a package that depends on virtual/mta, presumably because it needs to send emails. Had you installed qmail using portage, the virtual/mta dep would have been satisfied by it, and nullmailer would not have been installed in the first place. However, you didn't do that, and so portage had no idea qmail was installed. A possible workaround would be to add mail-mta/netqmail to package.provided on your system. However, there's still no guarantee that your custom-built qmail software will work with other packages provided by Gentoo. Regarding your accusations: Gentoo developers cannot anticipate every possible thing you might do on your system, especially when you start installing custom programs in paths that are traditionally managed by our package manager. Using portage you can customize your system extensively, without needing to custom build your own software. If that's not good enough for you, go build a Linux from Scratch system and enjoy the lack of any package management or support whatsoever. I was also surprised to see the installation of a mta in an emerge update, so I masked virtual/mta to see why this dependency was pulled. It turns out that app-crypt/gnupg depends on virtual/mta since version 2.2.6. Now the question is, why does gpg need a mta?
