Hi,
if you never heard of /var/adm/update-scripts and your packages are not using it, you can ignore this.
In the good old times, libzypp wasn't able to handle RPM %posttrans scripts correct (which shows, that the old times were not always good ;) ). A workaround for this was to use update-scripts.
Since some years (SLE12) libzypp supports now RPM %posttrans scripts, so there is no need anymore to use update-scripts as workaround. The maintainers would like now to remove support of /var/adm/update-scripts from libzypp. Why? Because it creates a lot of additional efford for everybody. You have to implement everything twice (once for RPM, once for libzypp) and test everything twice. And since this is so much work, a lot of developer don't do that, which leads to bugs depending on how you install or update a package.
So, if you use update-scripts as RPM %posttrans replacement, please change your spec files to only use RPM %posttrans scripts.
If you need update-scripts for something else and you cannot remove it, please speak up now. But you should have a good answer to "what does happen if your package is installed or updated with plain rpm?".
Thorsten
Thorsten Kukuk píše v St 22. 11. 2017 v 10:01 +0100:
Hi,
if you never heard of /var/adm/update-scripts and your packages are not using it, you can ignore this.
In the good old times, libzypp wasn't able to handle RPM %posttrans scripts correct (which shows, that the old times were not always good ;) ). A workaround for this was to use update-scripts.
Since some years (SLE12) libzypp supports now RPM %posttrans scripts, so there is no need anymore to use update-scripts as workaround. The maintainers would like now to remove support of /var/adm/update- scripts from libzypp. Why? Because it creates a lot of additional efford for everybody. You have to implement everything twice (once for RPM, once for libzypp) and test everything twice. And since this is so much work, a lot of developer don't do that, which leads to bugs depending on how you install or update a package.
So, if you use update-scripts as RPM %posttrans replacement, please change your spec files to only use RPM %posttrans scripts.
If you need update-scripts for something else and you cannot remove it, please speak up now. But you should have a good answer to "what does happen if your package is installed or updated with plain rpm?".
Great plan, I have but one question, could we first implement rpmlint error and fail all the current packages that are using it?
It is just one sure way to get them all fixed quickly.
Also overall I think most installations will have it from texlive and some fonts.
Cheers
Tom
On Nov 22 2017, Tomas Chvatal tchvatal@suse.cz wrote:
Also overall I think most installations will have it from texlive and some fonts.
The latter will be fixed automatically once the updated fontpackages package has hit Factory.
Andreas.
On Wed, Nov 22, 2017 at 12:26:15PM +0100, Andreas Schwab wrote:
On Nov 22 2017, Tomas Chvatal tchvatal@suse.cz wrote:
Also overall I think most installations will have it from texlive and some fonts.
The latter will be fixed automatically once the updated fontpackages package has hit Factory.
Correct. Nevertheless I would keep current rpm-macros.fonts-config for older distros.
Petr
On Wed, 22 Nov 2017 12:49:02 +0100, Petr Gajdos wrote:
On Wed, Nov 22, 2017 at 12:26:15PM +0100, Andreas Schwab wrote:
On Nov 22 2017, Tomas Chvatal tchvatal@suse.cz wrote:
Also overall I think most installations will have it from texlive and some fonts.
The latter will be fixed automatically once the updated fontpackages package has hit Factory.
Correct. Nevertheless I would keep current rpm-macros.fonts-config for older distros.
The old behavior is needed only for SLE11, right? Just for confirmation.
thanks,
Takashi
On Wed, Nov 22, 2017 at 12:58:39PM +0100, Takashi Iwai wrote:
The old behavior is needed only for SLE11, right? Just for confirmation.
I guess all the needed information is here: https://bugzilla.suse.com/show_bug.cgi?id=773575 The bug was for 12.3. Unfortunately I do not remember the details. Should I dive into?
Petr
On Wed, 22 Nov 2017 13:16:15 +0100, Petr Gajdos wrote:
On Wed, Nov 22, 2017 at 12:58:39PM +0100, Takashi Iwai wrote:
The old behavior is needed only for SLE11, right? Just for confirmation.
I guess all the needed information is here: https://bugzilla.suse.com/show_bug.cgi?id=773575 The bug was for 12.3. Unfortunately I do not remember the details. Should I dive into?
Not really. I just wanted to know the impact.
Takashi
On Wed, Nov 22, Tomas Chvatal wrote:
Great plan, I have but one question, could we first implement rpmlint error and fail all the current packages that are using it?
rpmlint check is not posssible: if you exclude texlive, the packages don't package the update-script, but create it on the fly (only look at all the font packages).
Else we have: - fetchmsttfonts => no idea why this is using update-scripts, doesn't make any sense to me. - mathgl => texlive, and broken. Does not work since the name of the script is wrong.
- nrpe => Misuse of the /var/adm/update-scripts directory as tmp directory, package needs to be fixed.
- nsca => Same broken code as nrpe
- texlive* => contains a lot of entries, where the script name does not match the package name, which means broken, will not work.
Conclusion: except for the font packages, the usage of update-scripts is always wrong, broken or misused. another reason to get ride of this hack rather today than tomorrow.
If anything else is using update-scripts, it is very well hidden in the code.
Thorsten
-
Andreas Schwab -
Petr Gajdos -
Takashi Iwai -
Thorsten Kukuk -
Tomas Chvatal