Re: [RFC PATCH] fast-export, fast-import: Let tags specify an internal name

[Elijah Newren <newren@gmail.com>]

On Wed, Apr 21, 2021 at 9:36 AM Luke Shumaker <lukeshu@lukeshu.com> wrote:
>
> On Wed, 21 Apr 2021 02:03:57 -0600,
> Ævar Arnfjörð Bjarmason wrote:
> > On Tue, Apr 20 2021, Luke Shumaker wrote:
> >
> > > -static void handle_tag(const char *name, struct tag *tag)
> > > +static void handle_tag(const char *refname, struct tag *tag)
> > >  {
> > >     unsigned long size;
> > >     enum object_type type;
> > >     char *buf;
> > > -   const char *tagger, *tagger_end, *message;
> > > +   const char *refbasename;
> > > +   const char *tagname, *tagname_end, *tagger, *tagger_end, *message;
> >
> > Let's put the new "*tagname, *tagname_end" on its own line, the current
> > convention is to not conflate unrelated variable declarations on the
> > same line (as e.g. the existing "message" and "tagger" does.
>
> Ack.
>
> > >     size_t message_size = 0;
> > >     struct object *tagged;
> > >     int tagged_mark;
> > > @@ -800,6 +801,11 @@ static void handle_tag(const char *name, struct tag *tag)
> > >             message += 2;
> > >             message_size = strlen(message);
> > >     }
> > > +   tagname = memmem(buf, message ? message - buf : size, "\ntag ", 5);
> > > +   if (!tagname)
> > > +           die("malformed tag %s", oid_to_hex(&tag->object.oid));
> > > +   tagname += 5;
> > > +   tagname_end = strchrnul(tagname, '\n');
> >
> > So it's no longer possible to export a reporitory with a missing "tag"
> > entry in a tag? Maybe OK, but we have an escape hatch for it with fsck,
> > we don't need one here?
> >
> > In any case a test for it would be good to have.
>
> I hadn't realized that it was possible for a tag object to be missing
> the "tag" entry, I will fix that.
>
> I don't think it's worth adding an option to fast-import to make it
> possible to create such an object, but fast-export should be able to
> handle it (and emit it in the stream such that fast-import would
> create it with the "tag" entry").
>
> Yes, the whole patch needs tests.

fast-export already dies on missing author or missing committer in a
commit object, so there seems to be some precedence for just dying
instead of swallowing objects.  (Is a missing "tag" line in a tag more
common that missing "author"/"committer" in commit objects?)

If we do want to add an option to handle the missing entry, perhaps we
make an option similar to fast-export's --fake-missing-tagger?

> > > @@ -884,14 +890,19 @@ static void handle_tag(const char *name, struct tag *tag)
> > >
> > >     if (tagged->type == OBJ_TAG) {
> > >             printf("reset %s\nfrom %s\n\n",
> > > -                  name, oid_to_hex(&null_oid));
> > > +                  refname, oid_to_hex(&null_oid));
> > >     }
> > > -   skip_prefix(name, "refs/tags/", &name);
> > > -   printf("tag %s\n", name);
> > > +   refbasename = refname;
> > > +   skip_prefix(refbasename, "refs/tags/", &refbasename);
> > > +   printf("tag %s\n", refbasename);
> > >     if (mark_tags) {
> > >             mark_next_object(&tag->object);
> > >             printf("mark :%"PRIu32"\n", last_idnum);
> > >     }
> > > +   if ((size_t)(tagname_end - tagname) != strlen(refbasename) ||
> >
> > Would be more readable IMO to have a temporary variable for that
> > "tagname_end - tagname", then just cast that and use it here.
> >
> > > +       strncmp(tagname, refbasename, (size_t)(tagname_end - tagname)))
> >
> > and here.
>
> Ack.
>
> > > @@ -2803,6 +2803,13 @@ static void parse_new_tag(const char *arg)
> > >     read_next_command();
> > >     parse_mark();
> > >
> > > +   /* name ... */
> > > +   if (skip_prefix(command_buf.buf, "name ", &v)) {
> > > +           name = strdupa(v);
> > > +           read_next_command();
> > > +   } else
> > > +           name = NULL;
> > > +
> >
> > Skip this whole (stylistically incorrect, should have {}) and just
> > initialize it to NULL when you declare the variable?
>
> In my defense, the guideline has always been to match the local style,
> and in fast-import.c this is done without {} 8 times and with {} 3
> times :)
>
> --
> Happy hacking,
> ~ Luke Shumaker