[Libre-soc-dev] daily kan-ban update 16oct2020

Luke Kenneth Casson Leighton lkcl at lkcl.net
Sat Oct 17 03:25:07 BST 2020


On 10/17/20, Jacob Lifshay <programmerjake at gmail.com> wrote:
> today:
> added more instructions to power-instruction-analyzer, then got
> waylaid by finding out that the new 379MiB output*.json file is too
> big for git to use a diff from the previous file

do you recall the rule that i set and emphasised in around 2 to 3
places in HDL_workflow about not adding autogenerated output to git
repositories?

do you now see why, from long experience, i set that rule?

lesson learned, now? :)

the size of the autogenerated output is irrelevant, the compressed
size is irrelevant: autogenerated output when added to a git *source
code* control system makes it absolutely impossible to use git diff
in.any meaningful way.

so the next task, to recover the repository to a useful and
maintainable state, involves git history rewriting to erase the
autogenerated output from it.  the *full* rewrite is needed (not, "git
rm last copy") to erase *every* copy of the autogenerated output, and
the process can be found online (stackexchange).  usually it can be
found in amongst "accidental added private keys or passwords".

once that is done a force push has to be done, followed by notifying
everyone to pull from the new master branch (by going through a
process of using "git fetch" not pul, deleting the old master branch
name, and creating a new master).

the file has to go, and yes be hosted on e.g. ftp.libre-soc.org, which
is what it is for (hosting of large autogenerated output).

l.



More information about the Libre-soc-dev mailing list