[Vm-dev] weird git history

[Nicolas Cellier]

Le lun. 7 janv. 2019 à 15:01, Nicolas Cellier <
nicolas.cellier.aka.nice at gmail.com> a écrit :

> Hi all,
> In an obsolete working copy, I had some local changes to
> build.win32x86/thirdparty/Makefile.libssh2
>
>     tar x -f $(LIBSSH2ARCHIVE) -C $(THIRDPARTYDIR)
>     cd $(LIBSSH2DIR) \
>         && ./configure \
> --            --build=x86_64-w64-cygwin \
> ++            --build=i686-w64-cygwin \
>             --host=i686-w64-mingw32 \
>             --prefix='$(THIRDPARTYOUTDIR)' \
>             CFLAGS='$(THIRDPARTY_CFLAGS)' \
>
> If I go into current version:
>
> https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/build.linux32x86/third-party/Makefile.libssh2
>
> I don't see the x86_64 line, good, this must have been already fixed!
> But when I look at history thru github interface to know when and how, I
> see a very partial and truncated history:
>
> https://github.com/OpenSmalltalk/opensmalltalk-vm/commits/Cog/build.linux32x86/third-party/Makefile.libssh2
>
> And itseems like none of the commits referenced did remove this x86_64
> line, like if it never existed...
>
> If I go a little further in the past commit log, in early november (still
> thru github interface
> https://github.com/OpenSmalltalk/opensmalltalk-vm/commits/Cog?after=0fae80c28ee3ce2b3eb09cb31ec94f0a590893f7+174),
> select one of the commits and go tho this same file, then i see another
> unrelated short history:
>
> https://github.com/OpenSmalltalk/opensmalltalk-vm/commits/fd213d14072e1099ded3590d99dd36e03562bcfe/build.win32x86/third-party/Makefile.libssh2
>
> This time, the x86_64 matching line is there...
> This kind of unconventional rewrite of history makes me worry...
> If we can't rely on history, then what a VCS is good for?
>
> It's possible that the minheadless merge introduced that, so it would be
> good if more pairs of eyes could check that we did not loose anything
> important, especially for Pharo builds.
>
> Or maybe it is from this commit?
https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/b6a1085868d16b10a06fe2443be1bf1a38a45a91
in which case i guess that the Pharo stuff should be OK.

But in such case, couldn't we rather rebase works performed on remote repos
rather than merge?
Otherwise, I find it very hard to understand a disrupted non linear
history... Or is it just me?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20190107/d1e813f3/attachment.html>