py-python-poppler-qt4: rename to py-poppler-qt4 and use PyPI

[dliessi (Davide Liessi)]

I decided to rename py-python-poppler-qt4 and to use PyPI as source instead of GitHub (see the thread starting with ​this message).
I had to remove the entry about py-python-poppler-qt4 from py-graveyard because MacPorts cannot handle "replaced_by chains" (see the thread starting with ​this message).

The changes are meant to be applied all at once:

• the Portfile for the new port py-poppler-qt4 is Portfile-py-poppler-qt4;
• the py-python-poppler-qt4 Portfile must be replaced by Portfile-py-python-poppler-qt4;
• the frescobaldi Portfile must be patched with Portfile-frescobaldi.diff;
• the py-graveyard Portfile must be patched with Portfile-py-graveyard.diff.

[dliessi (Davide Liessi)]

new port py-poppler-qt4

[dliessi (Davide Liessi)]

replace py-python-poppler-qt4 Poppler

[dliessi (Davide Liessi)]

patch py-graveyard

[dliessi (Davide Liessi)]

(Oops! The description of Portfile-py-python-poppler-qt4 was meant to say "replace py-python-poppler-qt4 Portfile" and not "... Poppler".)

[petrrr]

Cc Me!

[dliessi (Davide Liessi)]

patch frescobaldi (updated)

[dliessi (Davide Liessi)]

I updated the patch for frescobaldi to also increase the revision. I initially thought that this wasn't necessary: frescobaldi keeps running as expected, since py-python-poppler-qt4 and py-poppler-qt4 install the same files in the same locations. However dependencies of installed ports are not read from the Portfiles but from the registry, so py-python-poppler-qt4 is still considered a dependency of frescobaldi and cannot be uninstalled.

[petrrr]

Yes, dependencies are recorded in the registry.

So, when you change them but do not revbump, the port remains registered with this old dependencies. In some cases I was said not to revbump after such changes as this would trigger rebuilds without any real change the installed files. However, I found that in some cases it is necessary, especially if you want to resolve dependency conflicts. Here, you definitely want that the entry in the registry changes.

BTW: In future, please avoid submitting combined files and provide unified diffs. This will help a lot. The Portfile for py-poppler-qt4 is not really new, it has a history we want to preserve and we are supposed to commit non-semantic format changes as separated changesets. So I cannot commit immediately based on your files (therefore the delay).

[dliessi (Davide Liessi)]

Replying to petr@…:

Here, you definitely want that the entry in the registry changes.

Ok.

BTW: In future, please avoid submitting combined files and provide unified diffs. This will help a lot. The Portfile for py-poppler-qt4 is not really new, it has a history we want to preserve and we are supposed to commit non-semantic format changes as separated changesets. So I cannot commit immediately based on your files (therefore the delay).

I'm sorry, I don't fully understand: what did I do wrong and what should I have done? I'll submit my changes again in the preferred format, if you can explain this to me.

[dliessi (Davide Liessi)]

To be more precise:

Replying to petr@…:

In future, please avoid submitting combined files and provide unified diffs.

I understand that (according to ​the guide) I should have submitted diff files for each one of py-python-poppler-qt4 and py-poppler-qt4, but...

The Portfile for py-poppler-qt4 is not really new, it has a history we want to preserve

... py-poppler-qt4 is a new file, although it has similar content to the old py-python-poppler-qt4; on the other hand the new py-python-poppler-qt4 is completely different from the old one, so I thought it was clearer to read the new portfile instead of a meaningless diff. How should I have prepared a "history-preserving" diff?

and we are supposed to commit non-semantic format changes as separated changesets.

I don't think I understand this part.

[dliessi (Davide Liessi)]

Replying to petr@…:

and we are supposed to commit non-semantic format changes as separated changesets.

I don't think I understand this part.

I saw your commit r132349 and understood. (I don't know what I was thinking about. :D)

Let me know if you need me to do anything to help.

[dliessi (Davide Liessi)]

New patch set: patch 1

[dliessi (Davide Liessi)]

New patch set: patch 2

[dliessi (Davide Liessi)]

New patch set: patch 3

[dliessi (Davide Liessi)]

Please discard all patches not starting with a number!

I uploaded a new patch set, each patch corresponds to a single commit and does only one (set of related) thing(s).

They must be applied to the root of the dports tree with patch -p0 < patchname.diff in the order of the numbers.

Here are the proposed commit messages:
01-tiny-whitespace-change.diff​: "py-python-poppler-qt4: whitespace change"
02-python-poppler-qt4-switch-to-PyPI.diff: "py-python-poppler-qt4: use PyPI"
03-rename-py-poppler-qt4.diff: "py-python-poppler-qt4: rename to py-poppler-qt4; update dependency in frescobaldi"

[petrrr]

Committed in diff 01 in r133787 and diff 02 in r133788, with small changes to livecheck.

[dliessi (Davide Liessi)]

Thanks!

Later I'll rebase diff 03 on the current status, so it will apply cleanly.

[petrrr]

Unfortunately there is one problem with the patch 02. The checksums are actually incorrect as they are relative to the github package not the pypi package, but the mirrored packages would not change. So the situation is similar to a stealth update. ;-/.

Corrected in r133804. The result might be that checksum fails if the tarball is fetched from a macport mirror. As this is only temporary, I won't handle this as a stealth update.

[petrrr]

The name change and associated changes were committed in r133805. Closing as fixed.

[dliessi (Davide Liessi)]

Thanks!