From: "René Scharfe" <l.s.r@web.de>
To: Dominik Maier <dmnk@google.com>, git@vger.kernel.org
Cc: "Junio C Hamano" <gitster@pobox.com>,
"Patrick Steinhardt" <ps@pks.im>, "Johannes Sixt" <j6t@kdbg.org>,
"Jean-Noël Avila" <jn.avila@free.fr>,
dmnk+artist@google.com
Subject: Re: [PATCH] archive: reject remote --add-file in callback
Date: Sat, 21 Nov 2026 16:15:29 +0100 [thread overview]
Message-ID: <389c4ebc-7c4e-4ded-a620-b6a63e3911a3@web.de> (raw)
In-Reply-To: <20261119192311.2685319-1-dmnk@google.com>
On 11/19/26 8:23 PM, Dominik Maier wrote:
> parse_archive_args() only checks "is_remote && args->extra_files.nr"
> after parse_options() returns. When git-upload-archive parses client
> arguments, add_file_cb() runs first and calls stat() on the requested
> path on the server filesystem, reporting "File not found" or "Not a
> regular file" before the "--remote" incompatibility check is reached
> (and succeeding altogether if "--no-add-file" follows).
Ugh, good find.
> Pass is_remote into add_file_cb() and reject "--add-file" and
> "--add-virtual-file" immediately before touching the filesystem.
The current code rejects the combination of --add-virtual-file and
--remote, but that's arguably also a bug -- there's no reason to
disallow it, since the option's value brings file name and content
to the remote end.
> Signed-off-by: Dominik Maier <dmnk@google.com>
> ---
> archive.c | 31 ++++++++++++++++++++++---------
> t/t5000-tar-tree.sh | 9 +++++++++
> 2 files changed, 31 insertions(+), 9 deletions(-)
>
> diff --git a/archive.c b/archive.c
> index 59790be986..b983209ce4 100644
> --- a/archive.c
> +++ b/archive.c
> @@ -559,15 +559,27 @@ static void extra_file_info_clear(void *util, const char *str UNUSED)
> free(info);
> }
>
> +struct add_file_data {
> + struct archiver_args *args;
> + const char **basep;
> + int is_remote;
> +};
> +
> static int add_file_cb(const struct option *opt, const char *arg, int unset)
> {
> - struct archiver_args *args = opt->value;
> - const char **basep = (const char **)opt->defval;
> - const char *base = *basep;
> + struct add_file_data *data = opt->value;
> + struct archiver_args *args = data->args;
> + const char *base = *data->basep;
> char *path;
> struct string_list_item *item;
> struct extra_file_info *info;
>
> + if (data->is_remote)
> + die(_("options '%s' and '%s' cannot be used together"),
> + !strcmp(opt->long_name, "add-virtual-file") ?
> + "--add-virtual-file" : "--add-file",
> + "--remote");
That also rejects --remote --add-file=foo --no-add-file, which we could
allow.
> +
> if (unset) {
> string_list_clear_func(&args->extra_files,
> extra_file_info_clear);
> @@ -645,6 +657,11 @@ static int parse_archive_args(int argc, const char **argv,
> int i;
> int list = 0;
> int worktree_attributes = 0;
> + struct add_file_data add_file_data = {
> + .args = args,
> + .basep = &base,
> + .is_remote = is_remote,
> + };
> struct option opts[] = {
> OPT_GROUP(""),
> OPT_STRING(0, "format", &format, N_("fmt"), N_("archive format")),
> @@ -653,20 +670,18 @@ static int parse_archive_args(int argc, const char **argv,
> {
> .type = OPTION_CALLBACK,
> .long_name = "add-file",
> - .value = args,
> + .value = &add_file_data,
> .argh = N_("file"),
> .help = N_("add untracked file to archive"),
> .callback = add_file_cb,
> - .defval = (intptr_t) &base,
> },
> {
> .type = OPTION_CALLBACK,
> .long_name = "add-virtual-file",
> - .value = args,
> + .value = &add_file_data,
> .argh = N_("path:content"),
> .help = N_("add untracked file to archive"),
> .callback = add_file_cb,
> - .defval = (intptr_t) &base,
> },
> OPT_STRING('o', "output", &output, N_("file"),
> N_("write the archive to this file")),
> @@ -702,8 +717,6 @@ static int parse_archive_args(int argc, const char **argv,
> die(_("the option '%s' requires '%s'"), "--exec", "--remote");
> if (output)
> die(_("Unexpected option --output"));
> - if (is_remote && args->extra_files.nr)
> - die(_("options '%s' and '%s' cannot be used together"), "--add-file", "--remote");
How about moving the stat(2) call here instead? It would still happen
before we start writing archive content, but after applying any
--no-add-file. We can refuse to do it if is_remote is set. The files
added using --add-file have content == NULL.
>
> if (!base)
> base = "";
> diff --git a/t/t5000-tar-tree.sh b/t/t5000-tar-tree.sh
> index 3ad600c793..930a9fcb29 100755
> --- a/t/t5000-tar-tree.sh
> +++ b/t/t5000-tar-tree.sh
> @@ -237,6 +237,15 @@ check_tar with_untracked2
> check_added with_untracked2 untracked one/untracked
> check_added with_untracked2 untracked two/untracked
>
> +test_expect_success 'git archive --remote rejects --add-file before stat' '
> + test_must_fail git archive --remote=. --add-file=no-such-file HEAD 2>err &&
> + test_grep "cannot be used together" err &&
> + test_grep ! "File not found" err &&
> + echo untracked >untracked &&
> + test_must_fail git archive --remote=. --add-file=untracked --no-add-file HEAD 2>err &&
> + test_grep "cannot be used together" err
> +'
> +
> test_expect_success 'git archive on large files' '
> test_config core.bigfilethreshold 1 &&
> git archive HEAD >b3.tar &&
prev parent reply other threads:[~2026-11-21 15:15 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-19 19:23 [PATCH] archive: reject remote --add-file in callback Dominik Maier
2026-11-21 15:15 ` René Scharfe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=389c4ebc-7c4e-4ded-a620-b6a63e3911a3@web.de \
--to=l.s.r@web.de \
--cc=dmnk+artist@google.com \
--cc=dmnk@google.com \
--cc=git@vger.kernel.org \
--cc=gitster@pobox.com \
--cc=j6t@kdbg.org \
--cc=jn.avila@free.fr \
--cc=ps@pks.im \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox