non-ASCII char in xroot error message

[belforte]

looks like recent xroot (4.12.3-llifpc2) insert (at least at times) a non ASCII character at the end of an error message, see screenshot below.
This breaks current code which reports and parse those messages.
Can it be avoided ?
While I only have an example from using CMS code , @makortel says that in this case it just prints xrootd message "as is"

Of course we can filter for non-ASCII on our side, but @klannon suggested to report this to you as well for a proper resolution. In this I am the current maintainer of a tool which relies on CMS framework reports, so am 2 software layers (CMSSW and CMS/WMCore) away from xroot, simply acting as messenger here, but I doubt I can be of further help.
Thanks

[simonmichal]

@belforte : thank you for reporting this issue and sorry for replaying with such a delay!

Could you you test with the latest release, I believe this should have been fixed in: 22d308a

[belforte]

thanks @simonmichal . Unfortunately I am stuck with what version of xroot is built with CMSSW. But this should be easy to check on your side, just try to open a non-existing file. In any case I believe you can close.

[amadio]

Commit 22d308a likely fixes this, but it looks like commit a20fe39 could have introduced a potential off by one error. The function where the message comes from is below.

https://github.com/xrootd/xrootd/blob/v5.5.1/src/XrdCms/XrdCmsCluster.cc#L2031-L2045

If n could be off by one here, it could cause an extra '\0' character to be shown in the output due to Sel.Resp.Data containing two NUL characters at the end instead of just one. More specifically, in line 2040 if n is less than the size of the buffer, it's incremented by one, which adds the extra NUL character (note the +1 that should already account for it the line above).