- crux-devel - lists.crux.nu

system/iso.git/initramfs/init: do not scroll errors off screen
by Steffen Nurpmeso 25 May '18

25 May '18

Hello. This happened to me while installing CRUX 3.4, first on a VM, because of restricted memory the initramfs could not become extracted, but tar tried anyway, and all those messages flying by scrolled the error off screen; finally init(1) could not become started, resulting in a misleading error message. What do you think about the attached patch to "fix" this? --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt)

1 0

CRUX 3.4 has been released!
by Fredrik 12 May '18

12 May '18

We are happy to finally announce the release of CRUX 3.4 [1]. The toolchain has been updated to include glibc 2.27, GCC 7.3.0 and binutils 2.29.1. CRUX 3.4 ships with a 4.14.40 installation-kernel and Xorg 7.7 with xorg-server 1.20.0. For more in depth changes, check out the release notes [2] and changelog [3] [1] https://crux.nu/Main/Download [2] https://crux.nu/Main/ReleaseNotes3-4 [3] https://crux.nu/gitweb/?p=system/iso.git;a=blob;f=ChangeLog;h=49894a0760cd6…

1 0

[mpup] [patch v2] Don't delete ports just because sync fails
by John Vogel 18 Jan '18

18 Jan '18

From e8bdd627910f3f30a1c775cf815126b142af828a Mon Sep 17 00:00:00 2001 From: John Vogel <jvogel4(a)stny.rr.com> Date: Thu, 18 Jan 2018 06:26:50 -0500 Subject: [PATCH v2] Don't delete ports just because sync fails Check the list file instead. If the port doesn't exist or if it is commented out, then delete it. --- mpup | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/mpup b/mpup index 7730f20..7ee1093 100755 --- a/mpup +++ b/mpup @@ -160,7 +160,7 @@ while [ "$(cat $TMP_DIR/.pids 2>/dev/null | wc -l)" != "0" ]; do done done -# remove ports that were deleted from mpup.lst +# remove ports that were deleted from LIST_FILE for PORT in $ROOT_DIR/*; do PORT=${PORT##*/} [ "$PORT" == "*" ] && continue @@ -172,8 +172,8 @@ for PORT in $ROOT_DIR/*; do [ -z "$f" ] && continue # use the basename for file f="$(basename $f)" - # delete files of a port which was not sync'ed (.checkouts.$PORT missing) - if [ ! -f "$TMP_DIR/.checkouts.$PORT" ]; then + # delete files for ports without valid entries in LIST_FILE + if ! grep -qE "^[[:space:]]*(httpup|rsync).*[[:space:]]+$PORT[[:space:]]*$" $LIST_FILE; then rm -f $ROOT_DIR/$PORT/$f >/dev/null 2>&1 echo " Delete: $PORT/$f" fi -- 2.15.1

1 0

[mpup] Don't delete ports just because sync fails
by John Vogel 18 Jan '18

18 Jan '18

From e4e3ebed3dee31638bbcc5f425c9e0e90ad6bc88 Mon Sep 17 00:00:00 2001 From: John Vogel <jvogel4(a)stny.rr.com> Date: Thu, 18 Jan 2018 06:26:50 -0500 Subject: [PATCH] Don't delete ports just because sync fails Check the list file instead. If the port doesn't exist or if it is commented out, then delete it. --- mpup | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/mpup b/mpup index 7730f20..cc6aa84 100755 --- a/mpup +++ b/mpup @@ -160,7 +160,7 @@ while [ "$(cat $TMP_DIR/.pids 2>/dev/null | wc -l)" != "0" ]; do done done -# remove ports that were deleted from mpup.lst +# remove ports that were deleted from LIST_FILE for PORT in $ROOT_DIR/*; do PORT=${PORT##*/} [ "$PORT" == "*" ] && continue @@ -172,8 +172,8 @@ for PORT in $ROOT_DIR/*; do [ -z "$f" ] && continue # use the basename for file f="$(basename $f)" - # delete files of a port which was not sync'ed (.checkouts.$PORT missing) - if [ ! -f "$TMP_DIR/.checkouts.$PORT" ]; then + # delete files of a port removed or commented out in LIST_FILE + if ! grep -qE "^[^#].*[[:space:]]+$PORT$" $LIST_FILE; then rm -f $ROOT_DIR/$PORT/$f >/dev/null 2>&1 echo " Delete: $PORT/$f" fi -- 2.15.1

1 0

Cleaning Up giflib / libungif
by David L. Craig 15 Jan '17

15 Jan '17

Starting back in 2007 it is recorded in the #crux archives that the expiration of the GIF patent had deprecated libungif as far as CRUX was to be concerned. Today there is still collision in the known ports 'verse and maybe it is time to do something about it. According to my research, the following are available: opt giflib 5.1.4-1 [null Depends On] nwe libungif 4.1.4-1 [null Depends On] syl libungif 4.1.4-1 [no Depends On] noggplz libungif-32 4.1.4-1 [no Depends On] The following depend on giflib: enlightenment efi 1.8.6-1 enlightenment efi-e19 1.15.3-1 6c37 emacs 25.1-1 timcow emacs 24.5-1 baguette emacs-gtk3 24.5-1 opt fontforge 20150824-1 j_v icedtea7 2.6.8-1 opt imlib2 1.4.9-1 kde4 kdelibs 4.14.5-1 kf5 khtml 5.30.0-1 jaeger kodi 16.1-1 baguette mtpaint 3.40-3 6c37 openimageio 1.7.10-1 df openscenegraph 3.4-4 bdfy qgis 1.7.4-1 romster simage 1.7.0-1 contrib sxiv 1.3.1-1 contrib xscreensaver 5.35-1 The following depend on libungif: syl emacs 24.4-1 stdevel fbida 2.07-1 romster imlib 1.9.15-2 romster scm-imlib2 cvs-1 vico wmaker-crm 0.95.0-crm-1 contrib xplanet 1.3.0-1 Nothing seems to depend on libungif-32. There are depends on conflicts where installation of both giflib and libungif can be attempted by prt-get depinst. Does it make sense to eradicate libungif ports? Does it make sense to turn libungif into a prt-get alias of giflib? -- <not cent from sell> May the LORD God bless you exceedingly abundantly! Dave_Craig______________________________________________ "So the universe is not quite as you thought it was. You'd better rearrange your beliefs, then. Because you certainly can't rearrange the universe." __--from_Nightfall_by_Asimov/Silverberg_________________

2 1

Request for GPG/PGP keys!
by Fredrik Rinnestam 13 Oct '16

13 Oct '16

(Yes i realize the irony of not attaching my own) [1] So crux 3.3 is approaching and with that release we will introduce signify[2] as a way to do checksum and signing. To start with, core, opt and xorg will be signed. And to allow all these ports to be signed, we must distribute the private key to all maintainers. As this key should be well protected we must distribute this securely. So please reply and attach your public keys if you maintain ports in either opt or xorg! The 3.3 branch of core is already signed and you can take a look here: https://crux.nu/gitweb/?p=ports/core.git;a=shortlog;h=refs/heads/3.3 [1] Email sent from work! I'll reply to this email with my pub key later. [2] https://crux.nu/Wiki/SignedPorts -- Meddelandet har kontrollerats mot virus samt skadligt inneh�ll av MailScanner och f�rmodas vara s�kert.

5 5

prt-get design issue
by Vitaly Sinilin 14 Aug '16

14 Aug '16

Hi, Although CRUX IMHO is much more flexible in terms of configuration than other distros I seem to face the limits of the freedom it provides ;-) Tl;dr: the root cause of all difficulties I underwent was prt-get peeking into the pkgmk.conf file and I would like to discuss a solution for this. 1) What I was trying to do is to have each CRUX box storing its packages in its own directory on an NFS server named after the hostname. So I changed pkgmk.conf like PKGMK_PACKAGE_DIR="/crux/packages/$HOSTNAME" And pkgmk seemed to work well. But prt-get didn't. It was calling pkgmk, pkgmk was building the package, but prt-get was unable to install it. It turned out that prt-get had its own opinion regarding directory where packages are stored. Long story short, prt-get uses popen(3) to evaluate pkgmk.conf and popen() passes the command to /bin/sh which is not a symlink to /bin/bash anymore (and $HOSTNAME is a bashism). Yes, such configuration works fine on CRUX 2.8. Yes, I know that I can just probably use `hostname` (I've never tried it though), but it's not the point. The point is that two independent interpreters (I mean pkgmk itself and prt-get) of pkgmk.conf exist and special care from developers is required to keep them in sync. 2) I usually use something like this in prt-get.conf of my boxes makecommand sudo -H -u pkgmk /usr/bin/fakeroot /usr/bin/pkgmk and PKGMK_PACKAGE_DIR is only writable by the pkgmk user. What I don't like in this configuration is that to build a package privately as an unprivileged user I have to use some other pkgmk.conf file, what is not really bad per se, but passing this file from command line every time is a pain and I have to use some one-line wrapper named like mypkgmk. Recently I realized that this approach can be simplified. /etc/pkgmk.conf could contain a user-friendly configuration like build everything in $PWD because it is mostly used from command-line and prt-get can refer to another pkgmk.conf file that noone else is using. So I changed my prt-get.conf to makecommand sudo -H -u pkgmk /usr/bin/fakeroot \ /usr/bin/pkgmk -cf /etc/pkgmk-system.conf and got the same problem as in the first story. As you've already understand prt-get used /etc/pkgmk-system.conf for building the package, while /etc/pkgmk.conf for installing. Conclusion You may say that my configurations are perverse, but on the other hand nothing in the CRUX documentation states that such configurations are illegal. So I would say that from the principle of least astonishment they should work. If you are allowed to use custom make command you can expect that there is no other depencencies to the default command (by the way, it is never mentioned in prt-get(8) and prt-get.conf(5) that prt-get is peeking into hardcoded(!) /etc/pkgmk.conf). Obvious solution for this flaw is some kind of interface between prt-get and pkgmk that will allow the latter to inform the former about the directory where it will/would place the package. As a proof-of-concept I added the -dc|--dump-config option into pkgmk and modified prt-get so it calls $makecommand -dc and analyses its output instead of /etc/pkgmk.conf. It fixed both my issues: 1) $HOSTNAME is now expanded by pkgmk itself because --dump-config dumps values as they are interpreted by pkgmk (not verbatim values from pkgmk.conf); 2) prt-get asks "pkgmk -cf /etc/pkgmk-system.conf" about its configuration with the -dc option and gets correct PKGMK_PACKAGE_DIR value specified in /etc/pkgmk-system.conf. As a nice side effect this approach allows source-ing of pkgmk.conf files. E.g. my /etc/pkgmk-system.conf is the following # # /etc/pkgmk-system.conf: pkgmk(8) configuration # . /etc/pkgmk.conf PKGMK_SOURCE_DIR="/crux/distfiles" PKGMK_PACKAGE_DIR="/crux/packages/$HOSTNAME" (The interpreter of prt-get-5.19 won't be able to figure out PKGMK_COMPRESSION_MODE value from this configuration file.) I know from this mailing list that some people live with their own tuned versions of CRUX tools and I will probably do too, but as long as the spoken approach doesn't break any backward compatibility I think it can be considered for inclusion into the offical CRUX tools. -- Vitaly Sinilin <vs(a)kp4.ru>

1 0

prt-get source code beautifying
by Vitaly Sinilin 01 Aug '16

01 Aug '16

Hi, What is the reason to reset cout flags and fill character before each line? It makes no difference to output on my system, but makes code harder to read. diff --git a/src/prtget.cpp b/src/prtget.cpp index f22214c..5d94880 100644 --- a/src/prtget.cpp +++ b/src/prtget.cpp @@ -875,7 +875,6 @@ void PrtGet::printFormattedDiffLine(const string& name, cout << name; cout.width( 20 ); - cout.fill( ' ' ); cout << versionInstalled; string locked = ""; @@ -883,7 +882,6 @@ void PrtGet::printFormattedDiffLine(const string& name, locked = "locked"; } cout.width( 20 ); - cout.fill( ' ' ); cout << versionPortsTree << locked << endl; } /*! @@ -2044,7 +2042,6 @@ void PrtGet::printDepsLevel(int indent, const Package* package) void PrtGet::dumpConfig() { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); cout.fill( ' ' ); @@ -2055,58 +2052,40 @@ void PrtGet::dumpConfig() if ( m_parser->isAlternateConfigGiven() ) { fName = m_parser->alternateConfigFile(); } - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Configuration file: " << fName << endl; } if (m_config->cacheFile() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Cache file: " << m_config->cacheFile() << endl; } if (m_config->makeCommand() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Make command file: " << m_config->makeCommand() << endl; } if (m_config->addCommand() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Add command: " << m_config->addCommand() << endl; } if (m_config->removeCommand() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Remove command: " << m_config->removeCommand() << endl; } if (m_config->runscriptCommand() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Runscript command: " << m_config->runscriptCommand() << endl; } - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Run scripts: " <<(m_config->runScripts() ? "yes" : "no" ) << endl; - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Keep higher version:" <<(m_config->preferHigher() ? "yes" : "no" ) << endl; - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Readme mode: "; switch (m_config->readmeMode()) { case Configuration::VERBOSE_README: @@ -2124,34 +2103,23 @@ void PrtGet::dumpConfig() cout << endl; if (m_config->logFilePattern() != "") { - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << "Log file: " << m_config->logFilePattern() << endl; } - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << " Write log: " << (m_config->writeLog() ? "yes" : "no" ) << endl; - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << " Append log: " <<(m_config->appendLog() ? "yes" : "no" ) << endl; cout << endl; - cout.setf( ios::left, ios::adjustfield ); + cout.width( 20 ); - cout.fill( ' ' ); cout << "Pkgmk settings:" << endl; - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << " Package dir: " << InstallTransaction::getPkgmkPackageDir() << endl; - cout.setf( ios::left, ios::adjustfield ); cout.width( 20 ); - cout.fill( ' ' ); cout << " Compression mode: " << InstallTransaction::getPkgmkCompressionMode() << endl; -- Vitaly Sinilin <vs(a)kp4.ru>

1 0

pkgutils reloaded
by Alan Mizrahi 07 May '16

07 May '16

You are aboard a transatlantic flight on a packed A380, and the chicken served for lunch was tainted. Of course, none of the pilots wanted the pasta, so there is no one left to guide the plane anymore! Fortunately, the last act of bravery from the last pilot was to start the autopilot computer before colapsing, and as everybody knows, the autopilot software of Airbus is based on CRUX (tm). But not all is roses: the engineers forgot to run revdep before the last update, so the computer is running but the software has to be rebuilt urgently, before the plane goes down! As the only passenger familiar with CRUX [1], suddenly the lives of everybody on the plane are in your hands! You locate the port and fire up pkgmk, but after 4 minutes the build process fails. You quickly find out it built correctly, but the last "rm -r $PKG/usr/share/info" command failed. You comment that line, and run pkgmk again. As everybody knows, pkgmk has to start the build process from the beginning, taking up 4 more minutes of valuable time. If only pkgmk had a "--pack-only" option, you could have used it to just generate the package without rebuilding, start the autopilot and become a hero. Of course, this option would be much better if pkgmk automatically kept the work directory whenever the build fails. Fear no more, as this patch [2] for pkgutils implements both features, and some more. This allow you to save time and CPU cycles for long builds that fail, by finishing the build process manually, and then creating the package easily, instead of starting from scratch, or creating the package by hand. 1. set PKGMK_KEEP_WORK="failed" in /etc/pkgmk.conf, or use --keep-work=failed 2. pkgmk 3. Oops, something went wrong. Fix it manually. Make sure everything is in $PKG. 4. pkgmk --pack-only Here are all the changes implemented by this patch: 1. creates missing source, package and work directories automatically 2. prevents the creation of empty packages 3. detects number of CPUs and sets MAKEFLAGS accordingly 4. makes command-line options take priority over pkgmk.conf options 5. adds a DEBUG variable to build debug versions of packages easily: sudo -u pkgmk DEBUG=1 fakeroot pkgmk 6. adds an option to keep work directory only if the build fails: /etc/pkgmk.conf: PKGMK_KEEP_WORK="failed" 7. adds --pack-only, to just pack files in $PKG without building I think 2 and 4 could be considered bugs. And we should also make curl the default download program (instead of wget). Any opinions on these changes? I also made a patch for prt-get, but I haven't the story for that one yet ;) Cheers! Alan [1] I admit this part is a bit far fetched, since CRUX is widely popular, but I like the drama. [2] http://www.mizrahi.com.ve/crux/pkgs/pkgutils/pkgutils-alan-8.patch

2 1

Elegant(ish) way to deal with multiple qt versions
by Predrag Ivanovic 17 Feb '16

17 Feb '16

Hi. There is a lot of whining on #crux that Qt5 port is broken (someone suggested to nuke Qt4 because 'that's the way Arch does it'.Heh.) Interestingly enough, upstream has a solution for just this situation (more than one Qt version). It is "Qt Chooser - Tool to quickly switch between multiple Qt installations."[1] There is a simple port attached to FS task I opened a while back, using (now ancient) version of it.[2] Usage is simple enough: -Qt4 is the default, so any port that depends only on it builds OOTB. -If an app depends on Qt5, or it can use either, one extra line is needed in the Pkgfile (two, if it uses cmake instead of configure and friends): ---- export QT_SELECT=qt5 export CMAKE_PREFIX_PATH=/usr/share/qt5 ---- I've been using it a while, and it JustWorks[tm] So, considering that Qt4 is nowhere near being dead, and there will be the need for both versions in opt for quite some time, I humbly suggest adding qtchooser to opt. [1]https://code.qt.io/cgit/qt/qtchooser.git/ [2]https://crux.nu/bugs/index.php?do=details&task_id=1056 Pedja

1 0