intra-hunk data corruption: what is possible and what is not?

[Samuel Wales]

in my /old/ version of magit, i have positively found at least one,
probably two, cases where intra-hunk operations cause corruption. i
doubt anybody uses this version, and i don't know if i have notes that
say what to do to repro it.

it is not clear to me whether staging and unstaging alone can cause
the corruption, or whether killing is necessary. i have a note saying
that it doesn't really make sense for staging and unstaging alone to
do it, but i am not sure if that is so.

and there is the question of what kinds of git corruption are
possible. i do not know. i have seen that the actual working tree
data get corrupted. in one bug, individual line sequences can change.

in the other, i get duplication and strange errors in magit
operations, the only one i remember of which is that git is still
running. i think it is possible in principle that i have seen
corruption in git and not in git. i.e. in the latter case, the
corruption shows up in git diff or git log as you would expect, but it
isn't really a git issue. in the former, i am not sure.

i did find a bug report related to corruption at
https://github.com/magit/magit/issues/3182 and iirc they concluded
that it didn't exist, or if it did, it was a git bug. or similar.
idk if this is related to either bug that i found.

the first bug, line sequences, concerns e.g. marking a single line,
and staging and unstaging, and idk if killing or even resetting is
necessary to repro, but after these operations, sequence is incorrect.
it takes maybe 3 actions to do.


although i am not ready to upgrade to modern org yet for a few
reasons, one of which is my need for --- +++ headers, i am curious
about a few things. i guess i will try to make a list of my
questions:

1. what does modern org do against corruption of any type? is it
possible without e.g. fs corruption or user error?

2. what kinds of /git/ corruption are possible?

3. which ones do not show up in git-fsck?

4. have there ever been any other concerns about intra-hunk
operations causing corruption?

at some point i will likely upgrade, but i will not do so for some
time probably. these questions, if desired to answer, will give me an
idea of what to expect, or at least, help me understand magit and git
and their interaction a little better.

for reference, my source files are org-mode. the second set of buggy
behavior relates to a previous commit containing an org file with an
entry strangely inside another entry. different diff algorithms show
up later with different strange results.


On 11/20/22, Samuel Wales <samol...@gmail.com> wrote:
> great answer, thank you.
>
> for clarity, the REASON i use a paleolithic version of magit is that i
> need those headers. :) i'll try to stick with old magit for the time
> being; dunno if emacs upgrades will allow that.
>
>
> On 11/20/22, Jonas Bernoulli <jo...@bernoul.li> wrote:
>> Samuel Wales <samol...@gmail.com> writes:
>>
>>> i am currently using a version of emacs that is probably not supported
>>> by modern magit. i tried it, anyway.
>>
>> You'll need at least Emacs 25.1.
>>
>>> in any case, however, i have been using a paleolithic version of
>>> magit, because it shows --- +++ headers in the status buffer.
>>
>> Very old indeed.
>>
>>> at some
>>> point, magit stopped showing those. but i have searched recently, and
>>> maybe it can show them?
>>
>> No, that's not possible anymore. The information in those lines is
>> quite redundant. In the future it be become possible to show those
>> lines again, for the sake of doing the same as other tools (okay and
>> also to be able to copy-paste into those other tools). But this is
>> a very low priority, i.e., this might become possible again as a
>> side-effect of large changes, not because it is a goal in itself.
>>
>>> i found magit-insert-status-headers in what look like remarkably
>>> comprehensive online docs, but also a report that it is/was slow
>>> [which doesn't really make sense if it is for --- +++ headers].
>>
>> This function is unrelated. It is responsible for showing the
>> information at the very top of the status buffer, such as the current
>> branch and its upstream. And the latest tag--that's the part that can
>> be very slow in some repositories.
>>
>>> does modern magit support --- +++ headers?
>>>
>>> anything else i might like to know about modern magit? it seems a lot
>>> of development has occurred.
>>
>> Err...yes. Tons of new stuff. It generally dwims more.
>>
>>> it seems the package is still well received.
>>>
>>> thanks! :).
>>
>> Cheers,
>> Jonas
>>
>
>
> --
> The Kafka Pandemic
>
> A blog about science, health, human rights, and misopathy:
> https://thekafkapandemic.blogspot.com
>


--
The Kafka Pandemic

A blog about science, health, human rights, and misopathy:
https://thekafkapandemic.blogspot.com

[Samuel Wales]

i just remembered 2 more questions that prompted me to make the previous query.

1. is it possible to track corruption by keeping a copy of the
working tree or so, such that a program or script will detect
corruption of any type?

2. does such a script or program exist?

3. is there a stress tester for magit or git that would work on real
repos? or perhaps at least on fake ones? intra-hunk?