From nkipps at gmail.com Fri Feb 5 22:52:39 2021 From: nkipps at gmail.com (Nathaniel Kipps) Date: Fri, 5 Feb 2021 23:52:39 -0500 Subject: [crossfire] SVN history issues Message-ID: Recently I was rooting through SVN history, trying to trace some old facesets and PNGs, when I found a few anomalies. Maybe someone with more experience than me is able to determine: * Can these be fixed retroactively? * If so, is it easier to fix them in SVN or Git? The first issue is that at least one time in the past, a number of files were moved or renamed, but instead of tracking the change, it was treated as a large number of deletions and new files. The instance I found was r1487, in which mwedel renamed most or all of the face images. This was presumably done under CVS, and the broken history was passed into SVN. This is certainly not a catastrophic problem, as it's not too hard to continue tracing the history of the file, but it's certainly annoying. The second (and more severe) issue is that a lot of old PNGs have become corrupted. An example would again be r1487, where it appears that all the old images are broken, and all the new ones are fine. Thanks to some folks in IRC, we determined that the *older* images have been corrupted by insertion of extra CR symbols, as though something was trying to convert the file from LF to CR/LF. My best guess is that the PNGs were tagged wrong in CVS, then during the CVS to SVN conversion, the accidental CR/LF conversion was done on the PNGs. Supporting this theory is r13991, in which anmaster comments "Fix incorrect svn properties ... that could in worst case result in corruption." Anyone have any thoughts on whether fixing these is possible, and if so, how much effort might it require? --DraugTheWhopper From mwedel at sonic.net Mon Feb 8 23:20:35 2021 From: mwedel at sonic.net (Mark Wedel) Date: Mon, 8 Feb 2021 21:20:35 -0800 Subject: [crossfire] SVN history issues In-Reply-To: References: Message-ID: On 2/5/21 8:52 PM, Nathaniel Kipps wrote: > Recently I was rooting through SVN history, trying to trace some old > facesets and PNGs, when I found a few anomalies. Maybe someone with > more experience than me is able to determine: > * Can these be fixed retroactively? > * If so, is it easier to fix them in SVN or Git? You assessment below on the causes is likely correct (going back to CVS repo). I think at the time of the CVS->SVN conversion, the CVS repos were still active, with the idea being that if anyone wanted to see the earlier history, they could still go back to the CVS repo. But it looks like the CVS repos disappeared - going by how many years since, probably not unreasonable. It is also worth mentioning that the CVS repo on sourceforge was not the first repo. I sort of recall a CVS repo hosted elsewhere, and before that, it was an RCS repo on the maintainers (my) machine. I suspect the commit messages may be completely lost - I checked my local system to see if maybe I had a backup of the CVS stuff repo itself, but I couldn't find one. The images may be recoverable - this presumes that the initial import to CVS didn't result in corruption, but rather the CVS -> SVN conversion resulted in corruption. If the image in questions are in the actual crossfire-arch distribution, versions exist on sourceforge for download that predate the SVN conversion, so it may be possible to download one of those arch distributions and just copy over the pre-corrupted images into the current arch tree and recommit. > > The first issue is that at least one time in the past, a number of > files were moved or renamed, but instead of tracking the change, it > was treated as a large number of deletions and new files. The instance > I found was r1487, in which mwedel renamed most or all of the face > images. This was presumably done under CVS, and the broken history was > passed into SVN. This is certainly not a catastrophic problem, as it's > not too hard to continue tracing the history of the file, but it's > certainly annoying. > > The second (and more severe) issue is that a lot of old PNGs have > become corrupted. An example would again be r1487, where it appears > that all the old images are broken, and all the new ones are fine. > Thanks to some folks in IRC, we determined that the *older* images > have been corrupted by insertion of extra CR symbols, as though > something was trying to convert the file from LF to CR/LF. My best > guess is that the PNGs were tagged wrong in CVS, then during the CVS > to SVN conversion, the accidental CR/LF conversion was done on the > PNGs. Supporting this theory is r13991, in which anmaster comments > "Fix incorrect svn properties ... that could in worst case result in > corruption." > > Anyone have any thoughts on whether fixing these is possible, and if > so, how much effort might it require? > > --DraugTheWhopper > _______________________________________________ > crossfire mailing list > crossfire at metalforge.org > http://mailman.metalforge.org/mailman/listinfo/crossfire > From nkipps at gmail.com Wed Feb 17 20:56:29 2021 From: nkipps at gmail.com (Nathaniel Kipps) Date: Wed, 17 Feb 2021 21:56:29 -0500 Subject: [crossfire] SVN history issues In-Reply-To: References: Message-ID: On Tue, Feb 9, 2021 at 12:33 AM Mark Wedel wrote: > The images may be recoverable - this presumes that the initial import to CVS didn't result in corruption, but rather the CVS -> SVN conversion resulted in corruption. If the image in questions are in the actual crossfire-arch distribution, versions exist on sourceforge for download that predate the SVN conversion, so it may be possible to download one of those arch distributions and just copy over the pre-corrupted images into the current arch tree and recommit. I'm not terribly worried about the images being permanently lost. What I'm more interested in is if it's possible to "fix" the history so that it's easier to browse in the future. And, if so, is it easier to fix it while we're using SVN, or once we move to Git? --DraugTheWhopper From mwedel at sonic.net Thu Feb 18 00:20:42 2021 From: mwedel at sonic.net (Mark Wedel) Date: Wed, 17 Feb 2021 22:20:42 -0800 Subject: [crossfire] SVN history issues In-Reply-To: References: Message-ID: On 2/17/21 6:56 PM, Nathaniel Kipps wrote: > On Tue, Feb 9, 2021 at 12:33 AM Mark Wedel wrote: > >> The images may be recoverable - this presumes that the initial import to CVS didn't result in corruption, but rather the CVS -> SVN conversion resulted in corruption. If the image in questions are in the actual crossfire-arch distribution, versions exist on sourceforge for download that predate the SVN conversion, so it may be possible to download one of those arch distributions and just copy over the pre-corrupted images into the current arch tree and recommit. > > I'm not terribly worried about the images being permanently lost. What > I'm more interested in is if it's possible to "fix" the history so > that it's easier to browse in the future. And, if so, is it easier to > fix it while we're using SVN, or once we move to Git? Unless someone has a backup of the CVS repo, I think the CVS commits/rename information is now lost (I looked, and don't have a CVS backup).