Arch Linux User Repository

I have updated Sentry to v9.1.1. This update is noteworthy and may require some changes during install or upgrade if you're planning on moving Sentry to a different system.

First, I have cleaned up several dependencies. In particular, Redis is still required, but I've opted to move it to the "optdepends" declaration for those who are running Sentry on a system separate from their Redis server or may be running Sentry in a systemd-nspawn container (I'm doing the latter). If you're installing Sentry for the first time and ignore the installation instructions, you may discover sentry-celery fails to start. If so, this is probably the reason.

Second, LLVM has also been moved into the "optdepends" array as it may not be required for everyone. In particular, LLVM is only necessary if you require dsym parsing. For users authoring software in Xcode, compiling with LLVM, or writing Rust, you will likely require LLVM to be installed alongside your Sentry instance (i.e. LLVM will need to be installed on the same system). This is an improvement for users who may not have this requirement as LLVM can pull in several related libraries.

Finally, I've removed the package ncurses5-compat-libs from the dependencies as it appears it is no longer needed. I've done some preliminary searching through the code base to validate this decision and have installed Sentry on some test containers, and I believe this to be true. If you encounter issues, verify that you have ncurses (from the official Arch repos) installed first before trying to install the old ncurses5 compatibility library. If you already have it installed, you may remove it.

If you encounter other issues with this particular release, please post them here. I don't expect you will, but there is a possibility as several old dependencies have been removed and the PKGBUILD has seen a small amount of cleanup. The installation and upgrade instructions have also been cleaned up, improved, and old information removed.

Sentry updated to 9.1.0. Some manual intervention may be required:

In particular, during the update process, pacman may complain about the file "/opt/sentry/lib/python2.7/ntpath.pyc"; remove this before continuing.

I'll soon be exploring the possibility of changing this package to depend on Python 3.x rather than Python 2.7. At present, some of the libraries Sentry depends on will fail during the build process (BeautifulSoup 3.x has one or more unit tests that fail). I will try to keep this migration as painless as possible, but bear in mind that you may need to upgrade to the new interpreter if you haven't already. Python 2.7 will be deprecated at the start of 2020.

Sentry updated to 9.0.0. You may notice an error akin to:

duplicate key value violates unique constraint "sentry_environmentproject_project_id_29250c1307d3722b_uniq"

This appears to be part of Sentry's normal operation as per #8004 [1].

Be aware that building this PKGBUILD on a machine that isn't in sync with the Sentry host may potentially create a package where uWSGI expects differing versions of glibc or libicu*. Your build and Sentry hosts should always be either 1) updated or 2) in sync.

[1] https://github.com/getsentry/sentry/issues/8004

I'll release another pkgrel tonight (possibly) with a patch for Sentry's requirements as it appears upstream still has psycopg2>=2.6.0,<2.7.0[1]. I wasn't able to find an active ticket on their repository, either, so feel free to submit one.

[1] https://github.com/getsentry/sentry/blob/ee10875b2649a80e221a5df9a2db675e00d5c3e6/requirements-base.txt#L32

The sentry build fails with postgres 10.1

Collecting psycopg2<2.7.0,>=2.6.0 (from sentry==8.22.0)

    Using cached psycopg2-2.6.2.tar.gz
    Complete output from command python setup.py egg_info:
    running egg_info
    creating pip-egg-info/psycopg2.egg-info
    writing pip-egg-info/psycopg2.egg-info/PKG-INFO
    writing top-level names to pip-egg-info/psycopg2.egg-info/top_level.txt
    writing dependency_links to pip-egg-info/psycopg2.egg-info/dependency_links.txt

writing manifest file 'pip-egg-info/psycopg2.egg-info/SOURCES.txt'

Error: could not determine PostgreSQL version from '10.1'

Related ticket:

https://github.com/psycopg/psycopg2/issues/594

Updating to psycopg2==2.7.1 seems to fix this issue

Apologies for the nearly 2 week delay bumping Sentry. I received the notification, but promptly forgot following the Thanksgiving holiday due to some personal matters that came up.

Strange. My instance already had `citext` installed, but I a) don't remember installing it and b) my Sentry user isn't a superuser and doesn't have any attributes other than login. I must've installed it some time ago, but that seems unlikely. It explains why I didn't catch this issue.

One thing I might recommend instead is to use the PostgreSQL super user so you don't have to modify the Sentry user (potentially forgetting it has superuser):

$ sudo -u postgres psql "dbname=sentry"
sentry=# create extension if not exists citext;

Obviously, substituting "sentry" for the actual database name of your Sentry install.

If anyone else runs into this issue, I might update the postupgrade/postinstall with some notes but the cause should be self-explanatory even if the solution is not immediately obvious.

Thanks, Mitch! Good catch as always.

Hmm, getting an error when upgrading from 8.19.0 to 8.20.0:

django.db.utils.ProgrammingError: ProgrammingError('permission denied to create extension "citext"\nHINT: Must be superuser to create this extension.\n',)
SQL: CREATE EXTENSION IF NOT EXISTS citext

I'm guessing because the "sentry" user for my postgresql database isn't a superuser, so I made it one:
ALTER USER sentry WITH SUPERUSER;

After performing migrations, I changed the user back:
ALTER USER sentry WITH NOSUPERUSER;

If anyone's curious for the source of the issue, I've linked to the actual Sentry GitHub issue (https://github.com/getsentry/sentry/issues/5502), which can explain it better than I can :)

Thanks zancarius

Also, note that for smoothly copying and pasting from the issue resolution into the iPython interpreter, you'll need to remove empty lines that are at the same indentation level.

Edit: Since I forgot the AUR comments don't retain indentation level, I went ahead and fixed the formatting for the fix @mitchhentges linked below[1]. This should copy and paste just fine.

[1] https://gist.github.com/zancarius/33cd0ae7425198a16a63a0024ebf9644