git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] bundle-uri: drain remaining response on invalid bundle-uri lines

From
Justin Tobler <jltobler@gmail.com>
Date
Apr 10, 2026, 15:38 UTC
Message-ID
<adkOJLfxs8TNGRjr@denethor>
In-Reply-To
<adiZTBH_70nrpiHe@pks.im>
On 26/04/10 08:31AM, Patrick Steinhardt wrote:
Show 22 quoted lines
> On Wed, Apr 08, 2026 at 12:49:48PM -0500, Justin Tobler wrote:
> > On 26/04/08 10:58AM, Toon Claes wrote:
> > > Because bundle-URIs are optional by design, I believe the changes in
> > > this series are sufficient. Also, the series [2] takes a similar
> > > approach: have the client gracefully continue in case of misconfigured
> > > bundles.
> > 
> > I'm still largely of the opinion that a server-side fix should be
> > implemented first. Unless we really don't care that a server may
> > advertise invalid bundle-uri info to a client, making the client ignore
> > the error doesn't address the root of the problem. I don't see a good
> > reason why we would want servers to keep doing this anyways.
> > 
> > To be clear, I'm not against also making the client more resilient since
> > a "fixed" client may still try to talk to an older server that still
> > misbehaves though.
> 
> I think that addressing the client-side is a good first step, as we need
> to also be mindful that Git is not the only implementation used on the
> server side. So even if we fixed Git itself to not report garbage bundle
> URIs, other servers still very much might. So ensuring that clients can
> handle these gracefully is a good thing to do.

I think it is questionable for a Git server to be sending clients malformed bundle-uri configuration. Do other Git implementations on the server-side exhibit this same behavior? If so, or we reasonably think they could and just want to be safe, then I agree that adjusting clients first to ignore invalid bundle-uri configuration from the server is reasonable.

Generally, I'm of the mindset that when a server is sending malformed/garbage data that the client doesn't expect, the client should should be more strict and error out. In this case though, since there are known affected Git versions and bundle-uri is an optional feature to begin with, it probably doesn't hurt to be more permissive.

Show 5 quoted lines
> That being said, I also think that we should fix the server side.
> Whether that needs to be part of this patch series though is a different
> question. Based on the proposed patch you posted it seems to be trivial
> enough though, so maybe it's worth it to just add that in as a second
> patch.

Ya, my main concern was that a client-side fix would mask its root cause. As long as it gets addressed though it's fine. I think it would be worth adding to this series, but if not I'm happy to send a follow up patch to fix it too.

-Justin
Previous: Patrick SteinhardtNext: Toon Claes
Message 5 of 10 in “bundle-uri: drain remaining response on invalid bundle-uri lines”
  1. bundle-uri: drain remaining response on invalid bundle-uri linesToon Claes, Apr 8, 2026
  2. Patrick SteinhardtApr 8, 2026
  3. Justin ToblerApr 8, 2026
  4. Patrick SteinhardtApr 10, 2026
  5. Justin ToblerApr 10, 2026
  6. Toon ClaesJun 24, 2026
  7. 0/2 Fix fatal error in git-clone(1) when reading empty bundle-URIToon Claes, Jul 8, 2026
  8. 1/2 bundle-uri: drain remaining response on invalid bundle-uri linesToon Claes, Jul 8, 2026
  9. Junio C HamanoJul 8, 2026
  10. 2/2 bundle-uri: stop sending invalid bundle configurationToon Claes, Jul 8, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.