Once I discovered how git apply can take diff files (or patch files) as input, I stopped using git stash in favor of plain old files. Easier to list and browse the contents of prior edits, also you can grep the files as method of search. I’ve even found myself copying and editing the diffs before applying.
OskarS · 32d ago
That is a very neat trick, I agree.
I personally approaches stashes as undoable "clean up", and I never have anything really important that I want to save there. If I do have something like that, I just commit with a "WIP <some-descriptive-string>" message and don't push it, then a "git reset --mixed HEAD^" when I want to get back to it.
However, just FYI: you can "grep" your stashes really easily if you want to. just "git stash list -p" gives you the diffs for all the stashes, by default in "less" where you can search them, but you can pipe it to grep if you want. I somewhat frequently do that with "git log", if I want to know "when did this variable change?" or whatever, just "git log -p" to get the log with diffs in less, then search for whatever it was with a slash.
barbazoo · 33d ago
Oh that’s clever, I’ll try that out. Looks like you could just do a git diff > file.patch.
Neat.
johnrob · 33d ago
You’ll also want to familiarize with “git apply -3 <file name>”, for when a diff can’t be applied cleanly. It will try “harder” to merge (three way method) and if it still fails it invokes the conflict merge “UX”:
<<<<<<<<<
=========
>>>>>>>>>
smcameron · 32d ago
There's also Neil Brown's "wiggle" program for applying patches that don't apply.
although on debian based systems I think you can just "apt install wiggle"
johnisgood · 32d ago
What does "applying patches that don't apply" mean exactly?
I know about wiggle, but I have not used it, to be honest.
smcameron · 28d ago
It means that if you do "patch -p1 --dry-run < some.patch", and it complains that it doesn't apply, wiggle can sometimes apply it anyway, and also, if you do "patch -p1 < some.patch", and it partially applies but with rejected hunks, wiggle can try to apply the rejected hunks.
johannes1234321 · 33d ago
git diff an pipe works, but committing and then `git format-patch` can export multiple patches and then includes metadata (commit message, date, author, etc.) which can make reasoning about such files a lot easier. In a plain diff you only got filename as metadata.
d3ckard · 33d ago
Thank you, will try. Useful bit of knowledge.
RaoulP · 33d ago
That’s a great idea, and very timely for me.
teeray · 33d ago
Maybe slightly O/T, but has anyone found a decent way to `git send-email` with email hosts that demand OAuth? (looking at you Outlook and Gmail)
ozarker · 33d ago
I think you could set up postfix to smtp forward to those services. So it could handle the oauth2 and you wouldn’t need to configure your client
Exchange historically had a tendency to mangle emails sent through it (whitespace changes, line wrap, etc), which is obviously bad news for patchmails. I dunno if it's any better these days.
dmarinus · 33d ago
davmail supports smtp through outlook(365)
ndegruchy · 32d ago
Yeah, I used DAVMail with Emacs+MSMTP+MPOP+notmuch for ages. Works really well, the only occasional thing I had to do was reauthenticate the token, which pops up in a browser window.
palata · 32d ago
I like doing it with aerc [1]. It's even possible to use aerc in parallel to another email client. Just open aerc for git-related emails, and that's it!
At that point, why not just use Pijul or even Darcs?
smcameron · 19d ago
Because the codebase you're working on is on github?
And I think you may underestimate the power of stgit. You can manage thousands of patches concurrently, no problem. If you're a maintainer getting patches from loads of people all the time, this is valuable. stgit has it's origins in quilt, which in turn has its origins in Andrew Morton's patch scripts[1], and I know for a fact that Andrew Morton actually managed thousands of patches at a time for years in his work on the linux kernel, because I once sent him a patch against those scripts, and he complained it was slow because I used an O(n^2) algorithm, which worked fine with a handful of patches, and I asked him how many patches he had, and he told me a number that was multiple thousands, so this isn't a hypothetical example.
I personally approaches stashes as undoable "clean up", and I never have anything really important that I want to save there. If I do have something like that, I just commit with a "WIP <some-descriptive-string>" message and don't push it, then a "git reset --mixed HEAD^" when I want to get back to it.
However, just FYI: you can "grep" your stashes really easily if you want to. just "git stash list -p" gives you the diffs for all the stashes, by default in "less" where you can search them, but you can pipe it to grep if you want. I somewhat frequently do that with "git log", if I want to know "when did this variable change?" or whatever, just "git log -p" to get the log with diffs in less, then search for whatever it was with a slash.
Neat.
<<<<<<<<<
=========
>>>>>>>>>
https://github.com/neilbrown/wiggle
although on debian based systems I think you can just "apt install wiggle"
I know about wiggle, but I have not used it, to be honest.
git-send-email also has some quirks for Outlook which have been recently merged.
Exchange historically had a tendency to mangle emails sent through it (whitespace changes, line wrap, etc), which is obviously bad news for patchmails. I dunno if it's any better these days.
[1]: https://drewdevault.com/2022/07/25/Code-review-with-aerc.htm...
https://stacked-git.github.io
And I think you may underestimate the power of stgit. You can manage thousands of patches concurrently, no problem. If you're a maintainer getting patches from loads of people all the time, this is valuable. stgit has it's origins in quilt, which in turn has its origins in Andrew Morton's patch scripts[1], and I know for a fact that Andrew Morton actually managed thousands of patches at a time for years in his work on the linux kernel, because I once sent him a patch against those scripts, and he complained it was slow because I used an O(n^2) algorithm, which worked fine with a handful of patches, and I asked him how many patches he had, and he told me a number that was multiple thousands, so this isn't a hypothetical example.
[1] https://lwn.net/Articles/13518/
Then you don't need that message to be in a file-based inbox that is accessible from your git repo.
And in that case you are still likely going to have to copy and paste something to get the correct path.