On Sun, 2015-05-31 at 22:49 +0200, Lukas Fleischer wrote:
On Sun, 31 May 2015 at 22:09:19, Marcel Korpel wrote:
* Daniel Wallace <danielwallace@gtmanfred.com> (Sun, 31 May 2015 14:33:38 -0500):
The names usually end in 'hib' instead of humble bundle, standing for Humble Indie Bundle, but there are several other things out there that are similar to this.
Ah, now I understand. But those packages contain filenames with '://' in it, so I think this patch will just let those PKGBUILDs go through without those files being present, am I right?
Nevertheless, there are other cases I didn't think of, like https://aur.archlinux.org/packages/ttf-ms-win8/ where source files should be provided by the user. Providing zero-length dummy files looks like a solution, but isn't, as the checksums provided should be correct against those user-provided files, not the dummy ones.
I can't think of a solution to this at the moment, perhaps someone else?
Why are checksums an issue? You can use the checksum of the correct file. It doesn't match the checksum of the dummy file but I don't see how that is an issue (it is even good since the user immediately notices that something is wrong with the dummy file). Another possibility is to tell makepkg to skip the integrity check.
And *if* we go for solution 2, it should indeed be well-documented.
Best, Marcel
I am against dummy files and would even prefer dropping the patch in favor of a clean processing of files. Correct me if I am wrong but since the information is extracted from the .SRCINFO file, the package ttf-m-win8 should work just fine. The only problem is which files are delivered and which shell be downloaded. As things stand right now everything with " ://" or "lp:" in its filename is considered an URL and therefore the present of the file is not checked. This would potentially ignore cases where those files are omitted though not downloadable. However considering that this will help the vast majority where this schema fits, the minority of missing warnings are neglectable.