Junio C Hamano <gits...@pobox.com> writes: > Jeff King <p...@peff.net> writes: > >> I suspect that's more work because we'd need to refactor pretty.c a bit >> to make the right functionality available. But the end result would be >> much more maintainable. > > I actually think the entire codeflow of "find positions and length > of threeparts" using find_subpos() and then "copy the length bytes > starting position for C_{SUB,BODY,SIG,LINES,...}" must be rethought, > if the behavior of pretty.c::pretty_print_commit() is to be matched. > With the current code, %(contents:body) and other atoms that are > handled in ref-filter.c::grab_sub_body_contents() keep trailing > whitespaces on their lines with the current code that copies length > bytes starting the position using xmemdupz(). There need to be some > code that loses these trailing whiltespaces in the copied result. > > While I do not claim that refactoring and reusing code from pretty.c > is the only viable way forward, it is clear to me that a patch that > updates find_subpos() and changes nothing else falls short X-<.
I wonder if this would be a viable alternative (this is just a smoking-break hack without an attempt to think through corner cases---for example we need to make sure we work sensibly when the object does not have _any_ body past the header, but I do not think the original works well in that case, either). ref-filter.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-) diff --git a/ref-filter.c b/ref-filter.c index 1fc5e9970d..10f8fe15f5 100644 --- a/ref-filter.c +++ b/ref-filter.c @@ -949,13 +949,7 @@ static void find_subpos(const char *buf, unsigned long sz, const char **sig, unsigned long *siglen) { const char *eol; - /* skip past header until we hit empty line */ - while (*buf && *buf != '\n') { - eol = strchrnul(buf, '\n'); - if (*eol) - eol++; - buf = eol; - } + /* skip any empty lines */ while (*buf == '\n') buf++; @@ -1011,10 +1005,11 @@ static void append_lines(struct strbuf *out, const char *buf, unsigned long size } /* See grab_values */ -static void grab_sub_body_contents(struct atom_value *val, int deref, struct object *obj, void *buf, unsigned long sz) +static void grab_sub_body_contents(struct atom_value *val, int deref, struct object *obj, void *rawbuf, unsigned long sz) { int i; const char *subpos = NULL, *bodypos = NULL, *sigpos = NULL; + struct strbuf buf = STRBUF_INIT; unsigned long sublen = 0, bodylen = 0, nonsiglen = 0, siglen = 0; for (i = 0; i < used_atom_cnt; i++) { @@ -1030,11 +1025,18 @@ static void grab_sub_body_contents(struct atom_value *val, int deref, struct obj strcmp(name, "trailers") && !starts_with(name, "contents")) continue; - if (!subpos) - find_subpos(buf, sz, + if (!subpos) { + char *eoh = memmem(rawbuf, sz, "\n\n", 2); + eoh += 2; + sz -= eoh - (char *)rawbuf; + rawbuf = eoh; + strbuf_add(&buf, rawbuf, sz); + strbuf_stripspace(&buf, 0); + find_subpos(buf.buf, sz, &subpos, &sublen, &bodypos, &bodylen, &nonsiglen, &sigpos, &siglen); + } if (atom->u.contents.option == C_SUB) v->s = copy_subject(subpos, sublen); @@ -1060,8 +1062,9 @@ static void grab_sub_body_contents(struct atom_value *val, int deref, struct obj info.trailer_end - info.trailer_start); trailer_info_release(&info); } else if (atom->u.contents.option == C_BARE) - v->s = xstrdup(subpos); + v->s = xmemdupz(rawbuf, sz); } + strbuf_release(&buf); } /*