Reassigning to RPM; it turns out that the same error occurs if I run `rpm --test`. So there's a bug somewhere between the test transaction and the real transaction.

```
[root@75dffca6fbca noarch]# rpm --test -Uvh sgallagh-symlink-update-0-2.fc32.noarch.rpm
Verifying...                          ################################# [100%]
Preparing...                          ################################# [100%]
	file /opt/bug/package/etc/shells from install of sgallagh-symlink-update-0-2.fc32.noarch conflicts with file from package setup-2.13.6-2.fc32.noarch

```

Thanks for the nice reproducer!

Rpm currently only allows %pretrans presence to escape file conflicts on test upgrade in the very specific case that the conflicting file would get removed on the transaction anyway. Typically this would be an older version of the package itself, which naturally gets removed on an upgrade. Here, the conflict is on a file owned by setup, which isn't going anywhere. Of course the conflict is actually caused by an intermediate symlink that would go away in this transaction, but rpm doesn't see that.

Can't say offhand what, if anything, we can actually do about this (short of relaxing the %pretrans rule to cover more or less any conflict, which doesn't seem healthy) but at least the simple reproducer makes the case comprehensible.

(In reply to Panu Matilainen from comment #2)
> Thanks for the nice reproducer!
> 
> Rpm currently only allows %pretrans presence to escape file conflicts on
> test upgrade in the very specific case that the conflicting file would get
> removed on the transaction anyway. Typically this would be an older version
> of the package itself, which naturally gets removed on an upgrade. Here, the
> conflict is on a file owned by setup, which isn't going anywhere. Of course
> the conflict is actually caused by an intermediate symlink that would go
> away in this transaction, but rpm doesn't see that.
> 
> Can't say offhand what, if anything, we can actually do about this (short of
> relaxing the %pretrans rule to cover more or less any conflict, which
> doesn't seem healthy) but at least the simple reproducer makes the case
> comprehensible.

Well, the main problem here seems to be that RPM thinks it's removing /etc/shells when it's actually just removing the symlink to /etc/shells. That suggests to me that somewhere in the check, it's following the symlink instead of treating it as a file.

Yes, like I said it's the symlink to /etc.

Rpm simply treats a directory and a symlink to one as equal and doesn't generally consider whether it should follow something or not. This is a fundamental part of making various things work, but also causes these kind of problems.

I think if we add some comments in the %pretrans we see what's really going on.

## An actual update

# rpm -Uvh sgallagh-symlink-update2-0-4.el8.noarch.rpm
Verifying...                          ################################# [100%]
PRETRANS : sgallagh-symlink-update2-0-4.el8 
Preparing...                          ################################# [100%]
Updating / installing...
   1:sgallagh-symlink-update2-0-4.el8 ################################# [ 50%]
Cleaning up / removing...
   2:sgallagh-symlink-update2-0-3.el8 ################################# [100%]

## A test update

# rpm --test -Uvh sgallagh-symlink-update2-0-4.el8.noarch.rpm
Verifying...                          ################################# [100%]
Preparing...                          ################################# [100%]
        file /opt/bug/package/etc/shells from install of sgallagh-symlink-update2-0-4.el8.noarch conflicts with file from package setup-2.12.2-5.el8.noarch


As we see, during the --test, the %pretrans isn't happening.
And you don't want it to happen.
Running scripts during a --test is all sorts of bad.  It is no longer a test, it's the real thing, without the files being moved.

So, the whole %pretrans argument, if rpm is doing it correctly, seems to go out the window.
rpm is running %pretrans correctly.
But, during --test (which dnf calls) it isn't running %pretrans, which it shouldn't.


So, how to fix the problem.
A) have rpm treat symlinks to directories different.  But according to comment #4 that breaks other things.
B) don't have dnf call --test before it does updates.  That also leads to problems.  I remember the days of half done updates.
C) Something else

I'll be honest, I don't know what C is yet.  But it sure looks like that's what we need.

A promising work around is changing the link to point to a different directory.  That directory can be part of the rpm.
Using the example above, we start with 

#sgallagh-symlink-update3-0-1
/opt/bug/package/file1
/opt/bug/package/etc -> /etc

And we can safely update to the following

#sgallagh-symlink-update3-0-1
/opt/bug/package/file1
/opt/bug/package/etc -> /opt/bug/package/bundle-etc
/opt/bug/package/bundle-etc/
/opt/bug/package/bundle-etc/shells

That works for rpm -Uvh, rpm --test -Uvh, and dnf install or update

Created attachment 1700647 [details]
Specfile to show example, that says when pretrans is run

Created attachment 1700648 [details]
Specfile that shows possible work around

This message is a reminder that Fedora 32 is nearing its end of life.
Fedora will stop maintaining and issuing updates for Fedora 32 on 2021-05-25.
It is Fedora's policy to close all bug reports from releases that are no longer
maintained. At that time this bug will be closed as EOL if it remains open with a
Fedora 'version' of '32'.

Package Maintainer: If you wish for this bug to remain open because you
plan to fix it in a currently maintained version, simply change the 'version' 
to a later Fedora version.

Thank you for reporting this issue and we are sorry that we were not 
able to fix it before Fedora 32 is end of life. If you would still like 
to see this bug fixed and are able to reproduce it against a later version 
of Fedora, you are encouraged  change the 'version' to a later Fedora 
version prior this bug is closed as described in the policy above.

Although we aim to fix as many bugs as possible during every release's 
lifetime, sometimes those efforts are overtaken by events. Often a 
more recent Fedora release includes newer upstream software that fixes 
bugs or makes them obsolete.

Fedora 32 changed to end-of-life (EOL) status on 2021-05-25. Fedora 32 is
no longer maintained, which means that it will not receive any further
security or bug fix updates. As a result we are closing this bug.

If you can reproduce this bug against a currently maintained version of
Fedora please feel free to reopen this bug against that version. If you
are unable to reopen this bug, please file a new report against the
current release. If you experience problems, please add a comment to this
bug.

Thank you for reporting this bug and we are sorry it could not be fixed.