pacaur - an AUR helper that minimizes user interaction

X-dark · 2014-06-18 12:11:43

Nice. It means some work on our side, but that's a good reason to do some PKGBUILD cleanup.

While I understand the temporary aspect of the "insecure" option, why the "preview" option is not something you want to keep?

Spyhawk · 2014-06-18 15:51:00

X-dark wrote:

While I understand the temporary aspect of the "insecure" option, why the "preview" option is not something you want to keep?

Previewing before sourcing is unnecessary if you don't source PKGBUILDs at all. Note that you'll still have the option to view PKGBUILDs before installation.
I guess both options will be available until the AUR v4.0.0 release (see aur-dev mailing-list).

X-dark · 2014-06-18 15:54:09

Yes but broken dependencies can be solved that way. I know this is something that should be done by the maintainer but it may still be useful.
Anyway, I usually just pacaur -d any "not working as expected" PKGBUILD so I do not really mind.

Spyhawk · 2014-06-18 16:00:18

No, you won't be able to change dependencies at all since the dependency tree is solved through the extended RPC interface, which is fed by AUR metadata.
For not-working-as-expected cases, simply use pacaur/cower -d and makepkg.

X-dark · 2014-06-18 16:01:18

OK. I'll stick to pacaur -d then. Thanks for the info.

falconindy · 2014-06-18 17:11:41

Spyhawk wrote:

Pacaur 4.2.1 has been released and is now the stable version (see Changelog).
Before complaining about this new release, please read the following:
Since version 4.2, pacaur solves the dependency tree using the full secured extended RPC AUR interface. Packages which are missing AUR metadata may however not provide correct dependencies information and a warning will be displayed.
All packages uploaded to the AUR before 2014-05-27 18:35 UTC (AUR 3.0.0 release) are concerned and should be reuploaded by their respective maintainers. If you encounter such packages, please notify their maintainer about their missing AUR metadata status.
It is however possible to temporarily enable the old parsing dependency solver for such packages with the --insecure option, while the additional --preview option allows to preview PKGBUILDs before sourcing for dependency information.
Both of these options are temporary only and will be removed in the near future.

Does this mean I can remove -ii from cower and just rely on the .AURINFO data (or lack thereof)? I've already done most of it locally...

Spyhawk · 2014-06-19 08:16:10

falconindy wrote:

Does this mean I can remove -ii from cower and just rely on the .AURINFO data (or lack thereof)? I've already done most of it locally...

From what I've seen so far, the RPC interface v2 also provides depends information for packages that lack AUR metadata. It isn't always accurate (at least for complex PKGBUILDs), but might be "good enough" for most packages that don't have an .AURINFO file yet. However, the interface doesn't provide makedepends and optdepends info for such packages, so from a cower user point of view there is still value in the -ii option.

Also, pacaur doesn't rely on cower for dependency information anymore (although I'd like to rely on it again later in the future, once all packages have metadata), so the choice is entirely up to you here.

Last edited by Spyhawk (2014-06-19 09:08:02)

ceri · 2014-06-26 03:05:18

Bug:

pacaur -Syu

...

AUR Packages  (2):  package-query-1.4-1  yaourt-1.5-1  

:: Proceed with installation? [Y/n] 

...

:: Building yaourt package(s)...
==> Making package: yaourt 1.5-1 (Thu Jun 26 13:02:40 EST 2014)
==> Checking runtime dependencies...
==> Installing missing dependencies...
error: target not found: package-query>=1.4
==> ERROR: 'pacman' failed to install missing dependencies.

It tries to build yaourt first, even though the latest package-query is a dependency

Last edited by ceri (2014-06-26 04:37:24)

Spyhawk · 2014-06-26 07:43:56

ceri> Thx for the report. Please ensure you are running the latest version of pacaur, although I believe this is likely related to the 4.2 changes.

Could you also pastebin a debug output (bash -x pacaur -Sua)?

ceri · 2014-06-26 09:28:38

Spyhawk wrote:

ceri> Thx for the report. Please ensure you are running the latest version of pacaur, although I believe this is likely related to the 4.2 changes.
Could you also pastebin a debug output (bash -x pacaur -Sua)?

I'm running pacaur 4.2.3-1 which is the latest stable afaik.

I don't have the old version of package-query (nor a package file), so I'm not sure that debug output would still be helpful..?

Spyhawk · 2014-06-26 10:32:47

ceri wrote:

I don't have the old version of package-query (nor a package file), so I'm not sure that debug output would still be helpful..?

Yes please. I've tried to install the old yaourt 1.4 and package-query 1.3 and then upgrading with -Sua, but the packages installed in the correct order. I cannot reproduce the issue, and there might be something specific to your setup.

You can install the old version the "manual way" with makepkg:
- download the current PKGBUILDs, and change the version number (yaourt 1.4 -> 1.3, package-query 1.4 -> 1.3).
- modify the depends array for yaourt (package-query>=1.4 to package-query>=1.3).
- compile and install with makepkg -si --skipinteg (package query first, then yaourt).
- then you can upgrade again: "bash -x pacaur -Sua"

ceri · 2014-06-26 13:58:38

Spyhawk wrote:

ceri wrote:
I don't have the old version of package-query (nor a package file), so I'm not sure that debug output would still be helpful..?
Yes please. I've tried to install the old yaourt 1.4 and package-query 1.3 and then upgrading with -Sua, but the packages installed in the correct order. I cannot reproduce the issue, and there might be something specific to your setup.
You can install the old version the "manual way" with makepkg:
- download the current PKGBUILDs, and change the version number (yaourt 1.4 -> 1.3, package-query 1.4 -> 1.3).
- modify the depends array for yaourt (package-query>=1.4 to package-query>=1.3).
- compile and install with makepkg -si --skipinteg (package query first, then yaourt).
- then you can upgrade again: "bash -x pacaur -Sua"

Ok, here is the debug log:

http://pastebin.com/ug2mG2Fb

Edit: I was pipeing "bash -x ..." through tee but it seems to have lost a lot of the output. Here's the output from my terminal...

http://pastebin.com/tRxgGnRc

Last edited by ceri (2014-06-26 14:08:38)

Spyhawk · 2014-06-26 16:36:26

ceri> Unfortunately, it seems your debug output is incomplete (second link). My best guess is that your shell only retains the last 1000 lines or something, but the section of interest seems located beyond that limit.

Could you try to modify your shell parameters and try again? I've tried various settings on my machine, but the install goes as expected on my machine.

ceri · 2014-06-27 02:38:30

Spyhawk wrote:

ceri> Unfortunately, it seems your debug output is incomplete (second link). My best guess is that your shell only retains the last 1000 lines or something, but the section of interest seems located beyond that limit.
Could you try to modify your shell parameters and try again? I've tried various settings on my machine, but the install goes as expected on my machine.

Here's a full log:

http://pastebin.com/vvB9KdNX

I also had the same issue with another two packages:

AUR Packages  (2): android-sdk-r23-1  eclipse-android-23.0.0-1  
Repo Packages (1): libstdc++5-3.3.6-5  

Repo Download Size:   0.16 MiB
Repo Installed Size:  0.78 MiB

:: Proceed with installation? [Y/n] 

:: Retrieving package(s)...
:: View android-sdk PKGBUILD? [Y/n] 

:: android-sdk PKGBUILD viewed
:: View android-sdk.install script? [Y/n] 

:: android-sdk.install script viewed
:: View eclipse-android PKGBUILD? [Y/n] 

:: eclipse-android PKGBUILD viewed
:: Building eclipse-android package(s)...
==> Making package: eclipse-android 23.0.0-1 (Fri Jun 27 11:22:29 EST 2014)
==> Checking runtime dependencies...
==> Installing missing dependencies...
error: target not found: android-sdk>=r23
==> ERROR: 'pacman' failed to install missing dependencies.
:: eclipse-android cleaned
:: Building android-sdk package(s)..

Edit: as a sidenote, I'm using zsh.

Last edited by ceri (2014-06-27 02:40:21)

Spyhawk · 2014-06-27 07:52:39

ceri> Thx. I believe I found the cause of the bug: When the IgnorePkg variable in pacman.conf isn't empty, the order of AUR packages to be installed is messed up. This bug seems to have been in the code for a very long time... Will try to find a fix asap.

Edit: should now be fixed in master, and I'll release a fixed version soon. You can use pacaur-git in the meantime.

Last edited by Spyhawk (2014-06-27 08:56:51)

ceri · 2014-06-27 10:08:26

Spyhawk wrote:

ceri> Thx. I believe I found the cause of the bug: When the IgnorePkg variable in pacman.conf isn't empty, the order of AUR packages to be installed is messed up. This bug seems to have been in the code for a very long time... Will try to find a fix asap.
Edit: should now be fixed in master, and I'll release a fixed version soon. You can use pacaur-git in the meantime.

Thanks for the quick fix!

Last edited by ceri (2014-06-27 10:09:36)

krabat · 2014-07-03 15:39:38

As of 4.2.5 the -w switch doesn't seem to work while --downloadonly still does:

$ pacaur -Syuw 
  :: PKGDEST variable in /etc/makepkg.conf is unset and clean option is enabled

$ pacaur --downloadonly -Syu
  :: Synchronizing package databases...
  core is up to date
  [...]

No problems with any other operation like -Syu, -S or -R or if root is using the switch running pacman directly.
Both in makepkg.conf and pacman.conf there are only trivial changes such as packager or enabling multilib which shouldn't matter here, pacaur's config file is unchanged.
I tried to do some regression testing but didn't get PKGBUILD to work with older versions.

Any thoughts about this?

Spyhawk · 2014-07-03 16:11:18

krabat> Thx.. There a config setting check that prevents you to use -w if your cache directory isn't set, and that the clean option is enabled (this is the default on Arch). It's useless to compile AUR packages without installing them, and then deleting them from the hard drive.

Here, it seems the check only applied to the -w switch, and also wrongly applies for binary packages. I'll have a look shortly.

Edit: I removed the sanity check and replaced it by a note in the man page: if you're building without installing, ensure that PKGDEST is set, or that the clean option is disabled.
Edit2: New version 4.2.6 with the above fix is now released.

Last edited by Spyhawk (2014-07-06 17:33:13)

Markus00000 · 2014-07-16 10:09:12

I tried to install `linux-bfs` with pacaur and it aborted when user input was required. Not sure if handling user input is within the scope of pacaur, just wanted to let you know in case you haven't noticed.

Without waiting for input it aborted:

BFS cpu scheduler (SCHED_BFS) [Y/n/?] (NEW) aborted!

Building the package manually worked.

Spyhawk · 2014-07-16 10:18:43

Markus00000 wrote:

Not sure if handling user input is within the scope of pacaur, just wanted to let you know in case you haven't noticed.

Not within the scope, sorry.

Markus00000 · 2014-07-17 08:18:01

Okay. Is a one-line error message like "pacaur does not support packages that require user input. Try building it manually." within the scope? Of course, you'd have to be able to detect situations where the build process is aborted by waiting for user input.

After the "aborted!" line there were probably 20-30 more lines, therefore, it wasn't very obvious that the problem was actually pacaur (or let's say, the unsupported usage of pacaur).

Spyhawk · 2014-07-17 09:21:58

Markus00000 wrote:

Okay. Is a one-line error message like "pacaur does not support packages that require user input. Try building it manually." within the scope? Of course, you'd have to be able to detect situations where the build process is aborted by waiting for user input.

Not possible to do in an efficient manner afaik, but I might be able to add a note in the man page. Anyway, interactive PKGBUILDs shouldn't exist at all (and this is the primary reason I didn't and won't spend time trying to make them working).

Markus00000 · 2014-07-18 06:10:21

Understandable. Adding a note to the man page would be good. Both for those who actually read it and for you to tell the others to do so.

krabat · 2014-07-20 14:27:41

First of all thank you very much for your quick response.

The problem seems indeed to be fixed in 4.2.6, both for packages from AUR or regular repositories. Also, the advice you added to the man page works exactly as expected.

I'd like to add a question and some findings I stumbled upon recently.

As I understand it, both the configuration files in $XDG_CONFIG_DIRS and $XDG_CONFIG_HOME are handled by the shell's built-in 'source' command. Thus no problems affecting the one in $XDG_CONFIG_HOME have to be expected if /home resides on a dedicated partition mounted 'noexec'.
Is this correct?

If packages rely on VCS code and the corresponding client isn't installed yet, building will fail even if the client is stated as makedep as the packages from the repos are only installed after pacaur tries to fetch the VCS code:

$  pacaur -S libqtxdg-qt5-git
:: Package(s) libqtxdg-qt5-git not found in repositories, trying AUR...
:: resolving dependencies...
:: looking for inter-conflicts...

AUR Packages  (1): libqtxdg-qt5-git-latest  
Repo Packages (1): git-2.0.2-1  

Repo Download Size:    0.00 MiB
Repo Installed Size:  22.58 MiB

:: Proceed with installation? [Y/n] 

:: Retrieving package(s)...
:: View libqtxdg-qt5-git PKGBUILD? [Y/n] n
==> Making package: libqtxdg-qt5-git 0.5.3.17.gb3ccf52-1 (Sun Jul 20 15:36:59 CEST 2014)
==> Retrieving sources...
  -> Cloning libqtxdg git repo...
/usr/bin/makepkg: line 543: git: command not found
==> ERROR: Failure while downloading libqtxdg git repo
    Aborting...
:: Building libqtxdg-qt5-git package(s)...
==> Making package: libqtxdg-qt5-git 0.5.3.17.gb3ccf52-1 (Sun Jul 20 15:36:59 CEST 2014)
==> Checking runtime dependencies...
==> Checking buildtime dependencies...
==> Installing missing dependencies...
resolving dependencies...
looking for inter-conflicts...

Packages (1): git-2.0.2-1

Total Installed Size:   22.58 MiB

:: Proceed with installation? [Y/n] Y
(1/1) checking keys in keyring                                              [###########################################] 100%
(1/1) checking package integrity                                            [###########################################] 100%
(1/1) loading package files                                                 [###########################################] 100%
(1/1) checking for file conflicts                                           [###########################################] 100%
(1/1) checking available disk space                                         [###########################################] 100%
(1/1) installing git                                                        [###########################################] 100%
Optional dependencies for git
    tk: gitk and git gui
    perl-libwww: git svn
    perl-term-readkey: git svn
    perl-mime-tools: git send-email
    perl-net-smtp-ssl: git send-email TLS support
    perl-authen-sasl: git send-email TLS support
    python2: various helper scripts [installed]
    subversion: git svn
    cvsps: git cvsimport
    gnome-keyring: GNOME keyring credential helper
==> WARNING: Using existing src/ tree
==> Starting build()...
/tmp/pacaurtmp-peter/libqtxdg-qt5-git/PKGBUILD: line 24: cd: /tmp/pacaurtmp-peter/libqtxdg-qt5-git/src/libqtxdg: No such file or directory
==> ERROR: A failure occurred in build().
    Aborting...
:: libqtxdg-qt5-git cleaned
:: git is a new orphan package

Installation of the very same packages works smoothly if git was installed before. Problem does not affect building packages by 'makepkg -s'.

If pacaur is invoked as 'LANG=<locale != default locale> pacaur [...]' the output sometimes does switch to the default locale. This may prevent packages from being installed, apparently due to the question to do so being asked in the default locale's language but answered in the one corresponding to LANG. Example with default locale en_US.UTF-8:

$ LANG=de_DE.UTF-8 pacaur -S adminer 
:: Paket(e) adminer wurde in den Paketquellen nicht gefunden, suche im AUR...
:: Löse Abhängigkeiten auf...
:: no AUR metadata for adminer package
:: Suche nach Zwischenkonflikten...

AUR Pakete  (1):  adminer-4.1.0-2  
Repo Pakete (1):  php-5.5.14-1  

Repo Download Größe:     0.00 MiB
Repo Installierte Größe:  15.12 MiB

:: Fortfahren mit Installation? [J/n] 

:: Retrieving package(s)...
:: Betrachte adminer PKGBUILD? [J/n] n
:: View adminer.install script? [J/n] n
:: Building adminer package(s)...
==> Erstelle Paket: adminer 4.1.0-2 (So 20. Jul 15:52:45 CEST 2014)
==> Prüfe Laufzeit-Abhängigkeiten...
==> Installiere fehlende Abhängigkeiten... 
Löse Abhängigkeiten auf...
Suche nach Zwischenkonflikten...

Pakete (1): php-5.5.14-1

Gesamtgröße der zu installierenden Pakete: 15,12 MiB

:: Installation fortsetzen?  [J/n] J
(1/1) Prüfe Schlüssel im Schlüsselring                                      [###########################################] 100%
(1/1) Überprüfe Paket-Integrität                                            [###########################################] 100%
(1/1) Lade Paket-Dateien                                                    [###########################################] 100%
(1/1) Prüfe auf Dateikonflikte                                              [###########################################] 100%
(1/1) Überprüfe verfügbaren Festplattenspeicher                             [###########################################] 100%
(1/1) Installiere php                                                       [###########################################] 100%
==> Checking buildtime dependencies...
==> Retrieving sources...
  -> Downloading adminer-4.1.0.php...
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
  0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
100  378k  100  378k    0     0   262k      0  0:00:01  0:00:01 --:--:--  698k
==> Validating source files with md5sums...
    adminer-4.1.0.php ... Passed
==> Extracting sources...
==> Entering fakeroot environment...
==> Starting package()...
==> Tidying install...
  -> Purging unwanted files...
  -> Removing libtool files...
  -> Removing static library files...
  -> Compressing man and info pages...
  -> Stripping unneeded symbols from binaries and libraries...
==> Creating package "adminer"...
  -> Generating .PKGINFO file...
  -> Adding install file...
  -> Generating .MTREE file...
  -> Compressing package...
==> Leaving fakeroot environment.
==> Finished making: adminer 4.1.0-2 (Sun Jul 20 15:52:48 CEST 2014)
==> Installing package adminer with pacman -U...
loading packages...
resolving dependencies...
looking for inter-conflicts...

Packages (1): adminer-4.1.0-2

Total Installed Size:   0.37 MiB

:: Proceed with installation? [Y/n] J
==> WARNING: Failed to install built package(s).
:: adminer aufgeräumt
:: php ist eine neue verwaiste Paket

I'm not completely sure about the conditions when this takes place. But it seems that it always happens when installing an AUR package involves installing packages from the regular repositories (not from AUR). Re-running the command stated above will e. g. work flawlessly, probably as package php is available then.

If clean is set to false and 'pacaur -S some-aur-package' is invoked after 'pacaur -Sw some-aur-package' had been run before, the whole building and packaging procedure starts over, even if the package remained unchanged.

When building packages from VCS code there's a suffix foo-latest:

:: looking for inter-conflicts...

AUR Packages  (1): libqtxdg-qt5-git-latest

As there are no real packages called like so but rather all packages get installed without the suffix I wonder about its meaning and whether it's intentional in the first place.

The meaning of those messages

:: git is a new orphan package

isn't obvious to me either. I guess they are meant to express that those packages were installed as makedeps and are as such no longer required by other packages. If so I for one found this pretty misleading as "orphaned" is used to describe AUR packages which haven't got a maintainer any more.

LadFromWales · 2014-07-20 20:38:13

Since nvidia-full-beta was split into

pkgname=('nvidia-full-beta' 'nvidia-utils-full-beta' 'nvidia-libgl-full-beta' 'opencl-nvidia-full-beta')

Setting

_lib32=1

within the PKGBUILD does not result in the lib32 packages being created.

Building with makepkg poses no such problem.

Something I'm doing wrong or is this a bug?

Arch Linux

#426 2014-06-18 12:11:43

Re: pacaur - an AUR helper that minimizes user interaction

#427 2014-06-18 15:51:00

Re: pacaur - an AUR helper that minimizes user interaction

#428 2014-06-18 15:54:09

Re: pacaur - an AUR helper that minimizes user interaction

#429 2014-06-18 16:00:18

Re: pacaur - an AUR helper that minimizes user interaction

#430 2014-06-18 16:01:18

Re: pacaur - an AUR helper that minimizes user interaction

#431 2014-06-18 17:11:41

Re: pacaur - an AUR helper that minimizes user interaction

#432 2014-06-19 08:16:10

Re: pacaur - an AUR helper that minimizes user interaction

#433 2014-06-26 03:05:18

Re: pacaur - an AUR helper that minimizes user interaction

#434 2014-06-26 07:43:56

Re: pacaur - an AUR helper that minimizes user interaction

#435 2014-06-26 09:28:38

Re: pacaur - an AUR helper that minimizes user interaction

#436 2014-06-26 10:32:47

Re: pacaur - an AUR helper that minimizes user interaction

#437 2014-06-26 13:58:38

Re: pacaur - an AUR helper that minimizes user interaction

#438 2014-06-26 16:36:26

Re: pacaur - an AUR helper that minimizes user interaction

#439 2014-06-27 02:38:30

Re: pacaur - an AUR helper that minimizes user interaction

#440 2014-06-27 07:52:39

Re: pacaur - an AUR helper that minimizes user interaction

#441 2014-06-27 10:08:26

Re: pacaur - an AUR helper that minimizes user interaction

#442 2014-07-03 15:39:38

Re: pacaur - an AUR helper that minimizes user interaction

#443 2014-07-03 16:11:18

Re: pacaur - an AUR helper that minimizes user interaction

#444 2014-07-16 10:09:12

Re: pacaur - an AUR helper that minimizes user interaction

#445 2014-07-16 10:18:43

Re: pacaur - an AUR helper that minimizes user interaction

#446 2014-07-17 08:18:01

Re: pacaur - an AUR helper that minimizes user interaction

#447 2014-07-17 09:21:58

Re: pacaur - an AUR helper that minimizes user interaction

#448 2014-07-18 06:10:21

Re: pacaur - an AUR helper that minimizes user interaction

#449 2014-07-20 14:27:41

Re: pacaur - an AUR helper that minimizes user interaction

#450 2014-07-20 20:38:13

Re: pacaur - an AUR helper that minimizes user interaction

Board footer