Re: [PATCH v7 10/10] convert: add filter.<driver>.process option
- From
Lars Schneider <larsxschneider@gmail.com>
- Date
- Sep 12, 2016, 09:49 UTC
- Message-ID
- <10020380-76ED-4371-A0BA-59D07AF33CE0@gmail.com>
- In-Reply-To
- <20160910062919.GB11001@tb-raspi>
Show 20 quoted lines
> On 10 Sep 2016, at 08:29, Torsten Bögershausen <tboegi@web.de> wrote: > > On Thu, Sep 08, 2016 at 08:21:32PM +0200, larsxschneider@gmail.com wrote: > [] >> +packet: git> git-filter-client >> +packet: git> version=2 >> +packet: git> version=42 >> +packet: git> 0000 >> +packet: git< git-filter-server >> +packet: git< version=2 >> +packet: git> clean=true >> +packet: git> smudge=true >> +packet: git> not-yet-invented=true >> +packet: git> 0000 >> +packet: git< clean=true >> +packet: git< smudge=true >> +packet: git< 0000 > > It's probalby only me who has difficulties to distinguish > '>' from '<'.
I see what you mean. However, this format is used with "GIT_TRACE_PACKET" as well and therefore I would prefer to keep it.
Show 8 quoted lines
> packet: git> git-filter-client > packet: git> version=2 > packet: git> version=42 > packet: git> 0000 > packet: filter> git-filter-server > packet: filter> version=2 > > (Otherwise the dialoge description is nice)
Thanks!
Show 20 quoted lines
>> +------------------------ >> +Supported filter capabilities in version 2 are "clean" and >> +"smudge". >> + >> +Afterwards Git sends a list of "key=value" pairs terminated with >> +a flush packet. The list will contain at least the filter command >> +(based on the supported capabilities) and the pathname of the file >> +to filter relative to the repository root. Right after these packets >> +Git sends the content split in zero or more pkt-line packets and a >> +flush packet to terminate content. >> +------------------------ >> +packet: git> command=smudge\n >> +packet: git> pathname=path/testfile.dat\n > > How do we send pathnames the have '\n' ? > Not really recommended, but allowed. > And here I am a little bit lost, is each of the lines packed into > a pkt-line ? > command=smudge is packet as pkt-line and pathname= is packed into > another one ? (The we don't need the '\n' at all)
Every line is a dedicated packet. That's why '\n' in a path name would not be a problem as the receiver is expected to read the entire packet when parsing the value (and the receiver knows the packet length, too).
The '\n' at the end is required by the pkt-line format: "A non-binary line SHOULD BE terminated by an LF..." (see protocol-common.txt)
> Or go both lines into one pkt-line (thats what I think), then > we don't need the '\n' afther the pathname.
No (see above).
Show 12 quoted lines
> And in this case the pathname is always the last element, and a '\n' > may occur in the pathname, since we know the length of the packet > we know how long the pathname must be. > > > [...] >> >> +In case the filter cannot or does not want to process the content, > > Does not want ? > I can see something like "I read through the file, there is nothing > to do. So Git, I don't send anything back, you know where the file is.
That's right. Isn't that covered with "does not want"?
Show 29 quoted lines
>> +it is expected to respond with an "error" status. Depending on the >> +`filter.<driver>.required` flag Git will interpret that as error >> +but it will not stop or restart the filter process. >> +------------------------ >> +packet: git< status=error\n >> +packet: git< 0000 >> +------------------------ >> + >> +If the filter experiences an error during processing, then it can >> +send the status "error" after the content was (partially or >> +completely) sent. Depending on the `filter.<driver>.required` flag >> +Git will interpret that as error but it will not stop or restart the >> +filter process. >> +------------------------ >> +packet: git< status=success\n >> +packet: git< 0000 >> +packet: git< HALF_WRITTEN_ERRONEOUS_CONTENT >> +packet: git< 0000 >> +packet: git< status=error\n >> +packet: git< 0000 >> +------------------------ >> + >> +If the filter dies during the communication or does not adhere to >> +the protocol then Git will stop the filter process and restart it > > My personal comment: > When a filter is mis-behaving, Git should say so loud and clear, and > die(). > The filter process can be left running, so that it can be debugged.
In the current implementation Git would die already if the filter is "required". In this particular case we *could* die, too. However, I would prefer to keep it as is because I think the users of a "non-required" filter do not expect Git to die if there is *any* problem with the filter.
> Here I stopped the review for a moment, > I still think that Git shouldn't kill anything, because we loose > the ability to debug these processes.
Based on my experience debugging the filter in this state is hard anyways. I think a user would rather try to reproduce the problem and run Git with the "GIT_TRACE_PACKET" flag enabled for debugging.
Thanks, Lars