Re: [PATCH 1/6] ci: add a GitHub workflow to submit Coverity scans

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Sep 25, 2023 at 01:52:34PM +0200, Johannes Schindelin wrote:

> > You might want "--fail" or "--fail-with-body" here. I think any
> > server-side errors (like a missing or invalid token or project name)
> > will result in a 401.
> 
> Sadly, https://curl.se/docs/manpage.html#-f says this:
> 
> 	This method is not fail-safe and there are occasions where
> 	non-successful response codes slip through, especially when
> 	authentication is involved (response codes 401 and 407).
> 
> 401 is the precise case we're hitting when the token or the project name
> are incorrect.
>
> Having said that, I just tested with this particular host, and `curl -f`
> does fail with [exit code 22](https://curl.se/docs/manpage.html#22) as one
> would desire. So I will make that change.

The manpage is rather vague about what those "occasions" are, but I
think we are probably OK since we are not sending HTTP-level
credentials, according to:

  https://github.com/curl/curl/commit/cde5e35d9b046b224c64936c432d67c9de8bcc9e

At any rate, even if this did not catch every failure, I think we are
better off catching more rather than fewer.

> As to `--fail-with-body`: it is too new to use (it was [introduced in cURL
> v7.76.0](https://curl.se/docs/manpage.html#--fail-with-body) and Ubuntu
> v20.04 [comes with
> v7.68.0](https://packages.ubuntu.com/search?suite=focal&searchon=names&keywords=curl),
> i.e. is missing that option).
> 
> In any case, in my tests, `--fail-with-body` did not show anything more
> than `--fail` in this instance. Maybe for you it is different?

No, I don't think it's worth worrying about here. It could help with
debugging if their server returns something generic like a 500 code
along with a more detailed reason (we do something similar in
git-remote-curl to show failed bodies). But if it's at all more hassle
than typing "--with-body" it is not worth the effort.

> > I notice you put the "project" variable in the query string. Can it be
> > a --form, too, for symmetry?
> 
> The instructions at https://scan.coverity.com/projects/git/builds/new (in
> the "Automation" section) are very clear that `project` should be passed
> as a GET variable.
> 
> Even if using a POST variable would work, I'd rather stay with the
> officially-documented way.

That seems like good reasoning to me.

-Peff



[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux