Re: How to create the " [PATCH 0/5]" first email?
On Saturday 2012-09-15 19:08, Junio C Hamano wrote: > >If you plan to use "git send-email" to send the final results out, >you should consider "git send-email" as your "MUA" in the quoted >paragraph. And that will be very platform independent viewpoint to >see things from. > >"git format-patch -o my-series/ --cover-letter ..." would treat >"my-series/" directory as "MUA's drafts folder" and prepares the >messages you would want to send out, and you can proof-read and edit >the files in there before telling your "MUA" to send them out, with >"git send-email ... my-series/*.patch" or something. One can also send [0/n] with a normal MUA, and then use git send-email --in-reply-to '' commitrange It's not like 0/n has to be emitted at the same second 1/n is :) -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Monday, September 17, 2012 17:49:39 Junio C Hamano wrote: > "Philip Oakley" writes: > > I then applied it (using git am) to a temp branch to see what it > > produced, and could repeat the cycle until the patches looked right. > > That's another obvious and valid way to prepare your series. It all > depends on how comfortable you are to directly edit patches. Some > people fear it. Some don't. Some can do it with their eyes closed ;-). > > > However, when it came to creating the series, with comments, I > > couldn't see a way of having my comments within my local commits, but > > preparing a patch series that would properly include the '---' > > separator. > > An unofficial trick that works is to write the > > --- > > * This is an additional comment > > > yourself when running "git commit". That will be propagated to the > output from format-patch. You will have another "---" in front of > the diffstat, but nobody is hurt by that. One thing I have done is to add the additional comments I want with "git notes", then give the "--notes" option to format-patch or send-email. Unfortunately, this sticks the notes right into the commit message section, because the "--notes" option is actually a diff option, not something format-patch knows about, so you have to make sure to manually move it. But even so, I've found it a a nice way to track comments. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Tue, Sep 18, 2012 at 09:10:26PM +0100, Philip Oakley wrote: > From: "Junio C Hamano" > >have anything to do with DCO in any case? > > What's DCO an abbreviation of? Developer's Certificate of Origin. See SubmittingPatches. -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Tue, Sep 18, 2012 at 12:47:36PM -0700, Junio C Hamano wrote: > >> Jeff King writes: > >> > >> > But even without that, I still think format-patch is a reasonable time > >> > to do it. It is the time when I proof-read my commit message and patch > >> > in its final form, and think "do I really want to send this?". > >> > >> But it is not like "I cannot sign off because I think it is still > >> iffy". > > > > No, that is not the particular reason in my case, but I think I > > explained other reasons why "format-patch -s" is not a wrong workflow. > > Then I didn't read it. What does "do I really want to send this?" > have anything to do with DCO in any case? Because it is an excellent time to think about "am I willing and able to agree to the DCO?" As I said, for me personally working on git.git, that is not generally an issue. But I think it is perfectly reasonable for somebody to work and commit in isolation, and then only decide on the DCO during the sending phase (perhaps because they need to clear it with their company's legal department or some such). In other words, "it is iffy at the time of commit" might be exactly the reason for some people. If you are responding to my "that is not the particular reason in my case", I will paraphrase the reason I gave earlier: I find it annoying and pointless to type "-s" on every commit. We do not have commit.signoff, but we do have format.signoff. -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
From: "Junio C Hamano" have anything to do with DCO in any case? Junio, What's DCO an abbreviation of? Philip -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
Jeff King writes: > On Tue, Sep 18, 2012 at 12:11:58PM -0700, Junio C Hamano wrote: > >> Jeff King writes: >> >> > But even without that, I still think format-patch is a reasonable time >> > to do it. It is the time when I proof-read my commit message and patch >> > in its final form, and think "do I really want to send this?". >> >> But it is not like "I cannot sign off because I think it is still >> iffy". > > No, that is not the particular reason in my case, but I think I > explained other reasons why "format-patch -s" is not a wrong workflow. Then I didn't read it. What does "do I really want to send this?" have anything to do with DCO in any case? -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Tue, Sep 18, 2012 at 12:11:58PM -0700, Junio C Hamano wrote: > Jeff King writes: > > > But even without that, I still think format-patch is a reasonable time > > to do it. It is the time when I proof-read my commit message and patch > > in its final form, and think "do I really want to send this?". > > But it is not like "I cannot sign off because I think it is still > iffy". No, that is not the particular reason in my case, but I think I explained other reasons why "format-patch -s" is not a wrong workflow. > >> But your point still stands; "commit -s" will not see through that > >> official trick either ;-). > > > > Yes. :) > > Actually, no. "commit -s" does not have any need to see through it. > > ... hack hack hack ... > $ git commit -a -s > ... editor opens, you see your Sign-off at the end, with > ... the cursor sitting on the first line > ... edit the title, move to the line below the Sign-off, > ... and do the "---\n\n * comment" thing. > > And this survives "rebase -i" (but not "format-patch | am" for > obvious reasons). Yes, if your particular workflow is to signoff the very first time you commit. But it would not work for: ... hack hack hack ... $ git commit -a ... make a note after "---" ... ... hack hack hack ... ... OK, looks good, ready to signoff ... $ git commit --amend -s So it can work, but it is workflow dependent, and in general is a little flaky with the automagic signoff. You may want to signoff later for a variety of reasons, not the least of which is that you forgot to type "-s" the first time. > So I take it back. The time you do the "git commit" for the very > first time for this change that may need to be rerolled number of > times is the right time to say "-s". If you remember to type it. :) -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
Jeff King writes: > But even without that, I still think format-patch is a reasonable time > to do it. It is the time when I proof-read my commit message and patch > in its final form, and think "do I really want to send this?". But it is not like "I cannot sign off because I think it is still iffy". > seems to me like a reasonable time to make such a conscious decision to > signoff (or not). > >> But your point still stands; "commit -s" will not see through that >> official trick either ;-). > > Yes. :) Actually, no. "commit -s" does not have any need to see through it. ... hack hack hack ... $ git commit -a -s ... editor opens, you see your Sign-off at the end, with ... the cursor sitting on the first line ... edit the title, move to the line below the Sign-off, ... and do the "---\n\n * comment" thing. And this survives "rebase -i" (but not "format-patch | am" for obvious reasons). So I take it back. The time you do the "git commit" for the very first time for this change that may need to be rerolled number of times is the right time to say "-s". -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Mon, Sep 17, 2012 at 05:36:35PM -0700, Junio C Hamano wrote: > > But note that using "format-patch -s" will break; it puts the sign-off > > below the "---". > > I think "format-patch -s" is a workflow mistake in the first place. > You should be doing the sign-off the commit when you commit in the > first place. It is not like "I cannot sign off because I think it > is still iffy" or anything. It is mostly "I do not signoff because typing '-s' each time I commit is a pain". I know it is supposed to be a conscious decision, but I think it is a little silly. Since everything I do for git.git is going to be signed-off, either I will automate it, or I will end up typing it without thinking. Whenever I make a conscious decision, it is to _withhold_ signoff, since that is the exceptional case. But even without that, I still think format-patch is a reasonable time to do it. It is the time when I proof-read my commit message and patch in its final form, and think "do I really want to send this?". That seems to me like a reasonable time to make such a conscious decision to signoff (or not). > But your point still stands; "commit -s" will not see through that > official trick either ;-). Yes. :) -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
Jeff King writes: > On Mon, Sep 17, 2012 at 04:49:39PM -0700, Junio C Hamano wrote: > >> > However, when it came to creating the series, with comments, I >> > couldn't see a way of having my comments within my local commits, but >> > preparing a patch series that would properly include the '---' >> > separator. >> >> An unofficial trick that works is to write the >> >> --- >> >> * This is an additional comment >> >> >> yourself when running "git commit". That will be propagated to the >> output from format-patch. You will have another "---" in front of >> the diffstat, but nobody is hurt by that. > > But note that using "format-patch -s" will break; it puts the sign-off > below the "---". I think "format-patch -s" is a workflow mistake in the first place. You should be doing the sign-off the commit when you commit in the first place. It is not like "I cannot sign off because I think it is still iffy" or anything. But your point still stands; "commit -s" will not see through that official trick either ;-). >> But when doing a big series that deserves a cover letter [PATCH 0/n], >> you will use editor on the output from format-patch anyway, and I >> find it simpler to do the follow-on comments at that point myself. > > Me too (actually, I load it all into mutt and then comment on each as I > send them out, but it amounts to the same thing, seeing as how my MUA > just invokes $EDITOR when I edit a mail). -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
On Mon, Sep 17, 2012 at 04:49:39PM -0700, Junio C Hamano wrote: > > However, when it came to creating the series, with comments, I > > couldn't see a way of having my comments within my local commits, but > > preparing a patch series that would properly include the '---' > > separator. > > An unofficial trick that works is to write the > > --- > > * This is an additional comment > > > yourself when running "git commit". That will be propagated to the > output from format-patch. You will have another "---" in front of > the diffstat, but nobody is hurt by that. But note that using "format-patch -s" will break; it puts the sign-off below the "---". > But when doing a big series that deserves a cover letter [PATCH 0/n], > you will use editor on the output from format-patch anyway, and I > find it simpler to do the follow-on comments at that point myself. Me too (actually, I load it all into mutt and then comment on each as I send them out, but it amounts to the same thing, seeing as how my MUA just invokes $EDITOR when I edit a mail). -Peff -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
"Philip Oakley" writes: > I then applied it (using git am) to a temp branch to see what it > produced, and could repeat the cycle until the patches looked right. That's another obvious and valid way to prepare your series. It all depends on how comfortable you are to directly edit patches. Some people fear it. Some don't. Some can do it with their eyes closed ;-). > However, when it came to creating the series, with comments, I > couldn't see a way of having my comments within my local commits, but > preparing a patch series that would properly include the '---' > separator. An unofficial trick that works is to write the --- * This is an additional comment yourself when running "git commit". That will be propagated to the output from format-patch. You will have another "---" in front of the diffstat, but nobody is hurt by that. But when doing a big series that deserves a cover letter [PATCH 0/n], you will use editor on the output from format-patch anyway, and I find it simpler to do the follow-on comments at that point myself. Personal preferences vary, so whatever makes you feel comfortable with and works well for you is good. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
From: "Junio C Hamano" "Philip Oakley" writes: "Typically it will be placed in a MUA’s drafts folder, edited to add timely commentary that should not go in the changelog after the three dashes, and then sent as a message whose body, in our example, starts with "arch/arm config files were…". On the receiving end, readers can save interesting patches in a UNIX mailbox and apply them with git-am(1)." hides a multitude of implicit knowledge steps. Is there an extended description of what that would mean from a platform independent viewpoint? e.g. if the patches are separte files and an mbox is one consolidated file, how to get from one to the other so that 'it' can be sent by 'git send-mail'. If you plan to use "git send-email" to send the final results out, you should consider "git send-email" as your "MUA" in the quoted paragraph. And that will be very platform independent viewpoint to see things from. On git for windows (msysgit)n there were a couple of other steps I had to do https://git.wiki.kernel.org/index.php/MSysGit:UsingSendEmail as Outlooklook Express isn't really a suitable MUA ;-) [noted for other list readers] "git format-patch -o my-series/ --cover-letter ..." would treat "my-series/" directory as "MUA's drafts folder" and prepares the messages you would want to send out, and you can proof-read and edit the files in there before telling your "MUA" to send them out, with "git send-email ... my-series/*.patch" or something. I hadn't picked up from the man page that the --cover-letter would do the [PATCH 0/n] - should it? I'm also missing an understanding of the preparation stage where one tries to tidy up the various commit messages becaue they weren't explicit, specfic nor concise enough,... Many people usually do "rebase -i" until perfection and then a single final invocation of "format-patch". Of course, the "final" can and should further be proof-read and it is fine to do typofixes in the format-patch output files without going back to the commits before sending them out. I did an initial rebase to correct a few obvious mistakes (e.g. an extra file that had crept in), but then, after some false starts, used git format-patch pu --stdout > fix_Docs.patch to get a single file I could inspect and refine for both the commit messages and content. I then applied it (using git am) to a temp branch to see what it produced, and could repeat the cycle until the patches looked right. ...so I suspect that there is an implicit `git format-patch` <-> `git am` loop of sharpening the mbox patches before submission to the list that I'm missing. Has this described somewhere? -- However, when it came to creating the series, with comments, I couldn't see a way of having my comments within my local commits, but preparing a patch series that would properly include the '---' separator. Is there a way of getting format-patch to change some line break sequence (within the commit message) to the '---' three dashes patch break suitable for submission to the list? Philip -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: How to create the " [PATCH 0/5]" first email?
"Philip Oakley" writes: > "Typically it will be placed in a MUA’s drafts folder, edited to add > timely commentary that should not go in the changelog after the three > dashes, and then sent as a message whose body, in our example, starts > with "arch/arm config files were…". On the receiving end, readers can > save interesting patches in a UNIX mailbox and apply them with > git-am(1)." > > hides a multitude of implicit knowledge steps. Is there an extended > description of what that would mean from a platform independent > viewpoint? e.g. if the patches are separte files and an mbox is one > consolidated file, how to get from one to the other so that 'it' can be > sent by 'git send-mail'. If you plan to use "git send-email" to send the final results out, you should consider "git send-email" as your "MUA" in the quoted paragraph. And that will be very platform independent viewpoint to see things from. "git format-patch -o my-series/ --cover-letter ..." would treat "my-series/" directory as "MUA's drafts folder" and prepares the messages you would want to send out, and you can proof-read and edit the files in there before telling your "MUA" to send them out, with "git send-email ... my-series/*.patch" or something. > I'm also missing an understanding of the preparation stage where one > tries to tidy up the various commit messages becaue they weren't > explicit, specfic nor concise enough,... Many people usually do "rebase -i" until perfection and then a single final invocation of "format-patch". Of course, the "final" can and should further be proof-read and it is fine to do typofixes in the format-patch output files without going back to the commits before sending them out. > ...so I suspect that there is an > implicit `git format-patch` <-> `git am` loop of sharpening the mbox > patches before submission to the list that I'm missing. Has this > described somewhere? -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html