Re: [GSoC][PATCH v3] test: avoid pipes in git related commands for test

On Wed, Mar 21, 2018 at 2:11 PM, Junio C Hamano [off-list ref] wrote:

Pratik Karki [off-list ref] writes:

quoted

The changes in patch increased from v1 to v2 because I
got excited to work in Git codebase and I tried to
fix the exisiting problems as much as possible.
Hence, the large number of changes.

Eric understands why the scope was increased between the two; he
explained why it is not a good idea to increase the scope in the
middle, and I tend to agree with his reasoning.  The reason why the
scope was increased does not matter.

Thanks, Junio. I had just started writing a review of v3 when your
review arrived, and you covered every point I was going to make, thus
saved me the effort. I agree with everything in your review.

One additional comment, Pratik, is that this patch seems to be based
upon a slightly old version of the Git source code, thus does not
apply cleanly to present-day 'master'. Before re-rolling, update to
the latest Git and rebase your patch atop it.

quoted

-     PACK5=$( git rev-list --objects "$LIST" "$LI" "$ST" | \
-             git pack-objects test-5 ) &&
-     PACK6=$( (
+     git rev-list --objects "$LIST" "$LI" "$ST" >actual &&
+     PACK5=$( git pack-objects test-5 <actual ) &&
+     PACK6=$((

I thought that Eric already pointed out and explained why this
change to PACK6 is wrong in the previous round?

I probably should have been more explicit by naming PACK6 directly.
Comparing v3 against v2, I see that Pratik probably misunderstood my
comment, thinking that I was talking about losing the whitespace
inside PACK5=$(...); v2 dropped that whitespace and v3 restored it.

Pratik, dropping the unnecessary whitespace inside PACK5=$(...) is
fine (no complaint about that), but changing PACK6=$( (...) ) to
PACK6=$((...)) is outright incorrect as explained in [1].

[1]: https://public-inbox.org/git/CAPig+cTKkp6kpFcJfVV8W1ejCrCWQH33mHtgFUn+MpMgw5i1pA@mail.gmail.com/