Re: authz changes between 1.9 and 1.10
Branko Čibej wrote: >> https://cwiki.apache.org/confluence/x/7IjQBQ > > It is starting to improve. I'll be grateful to anyone who cares to take > a look to debug the syntax BNF. I appreciate the right-bracket rule change is in that domain. Could I nevertheless encourage you to jump ahead to writing the high-level rules, especially those that affected by the changes you propose in this thread relating to override/merge behaviours etc.? That's the part we need in order to review the proposed changes. -- - Julian
Re: authz changes between 1.9 and 1.10
On 05.12.2018 14:27, Julian Foad wrote: > Branko Čibej wrote: >>> https://cwiki.apache.org/confluence/x/7IjQBQ >> It is starting to improve. I'll be grateful to anyone who cares to take >> a look to debug the syntax BNF. > I appreciate the right-bracket rule change is in that domain. > > Could I nevertheless encourage you to jump ahead to writing the high-level > rules, especially those that affected by the changes you propose in this > thread relating to override/merge behaviours etc.? > > That's the part we need in order to review the proposed changes. One thing at a time. FWIW, the high-level rules are *already* documented in our wiki, as are changes from the previous authz incarnation (modulo bugs). I'll really just summarize them in a more user-friendly way. All known docs and issues are also linked from that page. In fact the documentation we have now, as well as test coverage, is much better than what we had before 1.10, when it was limited to half a chapter in The Book and to reading the code. It's kind of "fun" trying to document changes from undocumented behaviour. -- Brane
Re: authz changes between 1.9 and 1.10
On 05.12.2018 14:27, Julian Foad wrote: > Branko Čibej wrote: >>> https://cwiki.apache.org/confluence/x/7IjQBQ >> It is starting to improve. I'll be grateful to anyone who cares to take >> a look to debug the syntax BNF. > I appreciate the right-bracket rule change is in that domain. > > Could I nevertheless encourage you to jump ahead to writing the high-level > rules, especially those that affected by the changes you propose in this > thread relating to override/merge behaviours etc.? > > That's the part we need in order to review the proposed changes. Yes, but I'm asking for a review of the BNF to make sure that it doesn't contain silly bugs. -- Brane
Re: authz changes between 1.9 and 1.10
Branko Čibej wrote: > Yes, but I'm asking for a review of the BNF to make sure that it doesn't > contain silly bugs. At first glance, I saw what looks like a bug: it says a comment must have a non-white-space immediately after the '#'... I didn't review further. -- - Julian
Re: authz changes between 1.9 and 1.10
On 02.12.2018 17:28, Branko Čibej wrote: > On 02.12.2018 16:49, Branko Čibej wrote: > Seriously though: I started this document > https://cwiki.apache.org/confluence/x/7IjQBQ > Yes, it's empty. It will improve. It is starting to improve. I'll be grateful to anyone who cares to take a look to debug the syntax BNF. -- Brane
Re: authz changes between 1.9 and 1.10
On 05.12.2018 14:44, Julian Foad wrote: > Branko Čibej wrote: >> Yes, but I'm asking for a review of the BNF to make sure that it doesn't >> contain silly bugs. > At first glance, I saw what looks like a bug: it says a comment must have a > non-white-space immediately after the '#'... I didn't review further. The production used imply "optional". I already fixed that and how the comment definition reads: ::= "#" -- Brane
Re: authz changes between 1.9 and 1.10
On 05.12.2018 14:59, Branko Čibej wrote: > On 05.12.2018 14:44, Julian Foad wrote: >> Branko Čibej wrote: >>> Yes, but I'm asking for a review of the BNF to make sure that it doesn't >>> contain silly bugs. >> At first glance, I saw what looks like a bug: it says a comment must have a >> non-white-space immediately after the '#'... I didn't review further. > The production used imply "optional". I already fixed that and > how the comment definition reads: > > ::= "#" However /is/ wrong ... it implies that keys can't contain "#" or "[", where actually they can, they just can't start with one of those two. Whet they really can't contain is "=". I will have to check the code though ... -- Brane
Re: authz changes between 1.9 and 1.10
Branko Čibej wrote: > On 05.12.2018 14:44, Julian Foad wrote: > > Branko Čibej wrote: > >> Yes, but I'm asking for a review of the BNF to make sure that it doesn't > >> contain silly bugs. > > At first glance, I saw what looks like a bug: it says a comment must have a > > non-white-space immediately after the '#'... I didn't review further. > > The production used imply "optional". I already fixed that and > how the comment definition reads: > > ::= "#" The problem is isn't allowed to start with whitespace. -- - Julian
Re: authz changes between 1.9 and 1.10
On 05.12.2018 15:05, Julian Foad wrote: > Branko Čibej wrote: >> On 05.12.2018 14:44, Julian Foad wrote: >>> Branko Čibej wrote: Yes, but I'm asking for a review of the BNF to make sure that it doesn't contain silly bugs. >>> At first glance, I saw what looks like a bug: it says a comment must have a >>> non-white-space immediately after the '#'... I didn't review further. >> The production used imply "optional". I already fixed that and >> how the comment definition reads: >> >> ::= "#" > The problem is isn't allowed to start with whitespace. Eh, you're right; I redefined to fix that, which also fixes . -- Brane
