Closed
Bug 392004
Opened 19 years ago
Closed 14 years ago
Resurrect nsIScriptableInterfaceInfo & friends
Categories
(Core :: XPConnect, defect)
Core
XPConnect
Tracking
()
RESOLVED
WONTFIX
People
(Reporter: WeirdAl, Assigned: WeirdAl)
References
Details
(Keywords: regression)
Attachments
(3 files)
|
95.02 KB,
patch
|
peterv
:
review-
|
Details | Diff | Splinter Review |
|
10.36 KB,
text/plain
|
Details | |
|
26.82 KB,
patch
|
Details | Diff | Splinter Review |
The webservices extension, which primarily was about SOAP support, is no longer built by default. It has a scriptable interface info component, though, which is useful for my Verbosio project. I'd like to get that part of webservices brought back to life, either as another extension or as something XPConnect supports.
Thoughts?
Flags: blocking1.9?
| Assignee | ||
Comment 1•19 years ago
|
||
This code merely copies over the key parts of extensions/webservices needed for scriptable interface info. Taking a guess at appropriate reviewers based on bug 332174.
Issues I'm aware of:
* The jst-review simulacrum really doesn't like this patch. On the other hand, it's almost entirely copied from code that was already in the tree.
* This code shares UUID's with code still in extensions/webservices; no attempt has been made to remove code from there yet.
* I honestly do not know if this should be on by default. I'd like it to be (based on the fact it once was), but my use case is very weak: my design-by-contract library, for Verbosio. See http://weblogs.mozillazine.org/weirdal/archives/017721.html for details.
* This code may not belong in extensions/interfaceinfo, if other use cases exist. Possibilities include js/src/xpconnect/interfaceinfo or somewhere under XPCOM, in case PyXPCOM or JavaXPCOM want this.
I will not complain about an r- or sr- for any of the above reasons. I'm submitting this patch to get the ball rolling, and to determine if it can make Gecko 1.9.
Assignee: nobody → ajvincent
Status: NEW → ASSIGNED
Attachment #276730 -
Flags: superreview?(brendan)
Attachment #276730 -
Flags: review?(peterv)
| Assignee | ||
Comment 2•19 years ago
|
||
Upon request, I will submit a xpcshell-based testcase for this bug, using a frozen interface.
cc'ing sspeiche on this bug to make sure our stealing of webservices code won't conflict with your stealing of webservices code.
| Assignee | ||
Comment 4•19 years ago
|
||
aaronr: "borrowing" :)
It might simply be best to make sure XForms requires the interface info extension at build time, much like XForms requires schema-validation.
(In reply to comment #4)
> aaronr: "borrowing" :)
>
> It might simply be best to make sure XForms requires the interface info
> extension at build time, much like XForms requires schema-validation.
>
I don't believe that we'll have an issue with the interface. I don't see that we use it. The problem that might arise will be if we are 'borrowing' any of the same supporting classes :-)
Comment 6•19 years ago
|
||
Still wanted? I read your blog about reconsidering Verbosio in light of OpenKomodo. All else equal, I'd prefer not to restore this.
Was someone else who did not use SOAP feeling a need for this code after the SOAP removal?
/be
| Assignee | ||
Comment 7•19 years ago
|
||
(In reply to comment #6)
> Still wanted? I read your blog about reconsidering Verbosio in light of
> OpenKomodo. All else equal, I'd prefer not to restore this.
Would that include as an off-by-default extension? (Yes, I still want it, but my current use case is for debugging code only.)
> Was someone else who did not use SOAP feeling a need for this code after the
> SOAP removal?
Not that I have heard.
Comment 8•19 years ago
|
||
Comment on attachment 276730 [details] [diff] [review]
patch, v1
If it's only you, I'd rather not have it even as an off-by-default extension.
We are not likely to carry this into Mozilla 2, so you should cast a colder eye on this debugging code in that timeframe (2008 calendar year).
/be
Attachment #276730 -
Flags: superreview?(brendan)
What is this useful for? Why do we want it?
Comment 10•19 years ago
|
||
Scriptable interface info gives you the ability to introspect interface method parameters/return types... it can be useful for instance to provide autocomplete in a code editor.
Ok, not blocking on it though.
Flags: blocking1.9? → blocking1.9-
Whiteboard: [wanted-1.9]
| Assignee | ||
Comment 12•19 years ago
|
||
peterv, can you give me a timetable for review, please?
Comment 13•19 years ago
|
||
Comment on attachment 276730 [details] [diff] [review]
patch, v1
Please attach a diff of what exactly you changed and file a bug (with a CVS moves file) in mozilla.org/CVS: Copy that this one depends on.
Attachment #276730 -
Flags: review?(peterv) → review-
| Assignee | ||
Comment 14•19 years ago
|
||
peterv: I don't believe I changed anything, other than to add a C++ module file. I may not be able to generate a CVS moves file because I was using cvsutils... I'm not sure it has the capacity for that.
Comment 15•19 years ago
|
||
cvsmoves files are hand-generated files in a mozilla format.
| Assignee | ||
Comment 16•19 years ago
|
||
There's a serious discussion in #developers as to whether it's worth it:
<jst_home> WeirdAl: if there's useful history that's worth preserving, then fine, otherwise just mention where the files came from when you check in.
<jst_home> WeirdAl, reed: the reason I say that is that the cvs move scripts we have do wonky things in the hg repo
<WeirdAl> fwiw, extensions/webservices (the source of this mess) is not in Hg at present
The history does go back over five years...
Without further objection, I'll generate that move file... this weekend.
Comment 17•19 years ago
|
||
So we had a discussion about whether it's worth doing a cvs copy/move for these files on irc today, and I'm arguing it's not worth it. The reason I say that is primarily because of the fact that our cvs move scripts replay the checkins and thus show up as real checkins on hg.mozilla.org, and they're all bunched up and sometimes out of order (which is undetectable since the original checkin times are lost). What it also means is that a cvs move will pull history from the beginning of time into hg, which we don't want (due to repo size etc, see the blogs etc on this topic).
So my 2 cents are don't try to move files in cvs, just make sure to mention in the checkin comment that the files were moved from location x, and thus blame etc is still available at location x.
Comment 18•19 years ago
|
||
Ok, copy them, but I still would like a diff between the old files and the new ones.
| Assignee | ||
Comment 19•19 years ago
|
||
| Assignee | ||
Updated•19 years ago
|
Attachment #289196 -
Flags: review?(peterv)
Updated•18 years ago
|
Flags: wanted1.9+
Whiteboard: [wanted-1.9]
| Assignee | ||
Comment 20•18 years ago
|
||
missed 1.9, what about 1.9.1?
peterv: any chance of getting a review?
Flags: wanted1.9.1?
Comment 21•18 years ago
|
||
Not wanted, but if this is ready to go before freezes etc, feel free to land it.
Flags: wanted1.9.1? → wanted1.9.1-
| Assignee | ||
Comment 22•18 years ago
|
||
Comment on attachment 289196 [details] [diff] [review]
diff of copied files, per peterv's request
>--- extensions/webservices/interfaceinfo/src/Makefile.in Tue Dec 20 06:12:40 2005
>+++ extensions/interfaceinfo/Makefile.in Sat Nov 17 21:30:54 2007
>+EXTRA_DSO_LDOPTS = \
>+ $(XPCOM_GLUE_LDOPTS) \
>+ $(NSPR_LIBS) \
>+ $(MOZ_COMPONENT_LIBS) \
>+ $(NULL)
Turns out the XPCOM_GLUE_LDOPTS line breaks this patch.
Comment 23•18 years ago
|
||
XPCOM_GLUE_LDOPTS and MOZ_COMPONENT_LIBS don't go together. You probably want just XPCOM_GLUE_LDOPTS.
| Assignee | ||
Comment 24•18 years ago
|
||
Comment on attachment 276730 [details] [diff] [review]
patch, v1
>Index: extensions/interfaceinfo/Makefile.in
>+MOZILLA_INTERNAL_API = 1
>Index: extensions/interfaceinfo/iixprivate.h
>+#include "nsVariant.h"
>Index: extensions/interfaceinfo/nsScriptableInterfaceInfo.cpp
>+/* readonly attribute nsIVariant value; */
>+NS_IMETHODIMP
>+nsScriptableConstant::GetValue(nsIVariant * *aValue)
>+{
>+ nsVariant* variant = new nsVariant();
>+ if(!variant)
>+ {
>+ *aValue = nsnull;
>+ return NS_ERROR_OUT_OF_MEMORY;
>+ }
>+ *aValue = static_cast<nsIVariant*>(variant);
>+ NS_ADDREF(*aValue);
...
I found this nsVariant reference which required MOZILLA_INTERNAL_API. I'm thinking it probably shouldn't be that way, that we should use do_CreateInstance() and remove all other requirements for MOZILLA_INTERNAL_API.
| Assignee | ||
Comment 25•18 years ago
|
||
Is anyone actually willing to review this code for eventual check-in? Now that cvs has fallen to the mighty hg, any talk of cvs moves is pretty pointless.
| Assignee | ||
Comment 26•15 years ago
|
||
Let me throw an alternate idea out there. What parts of nsIInterfaceInfo could be made scriptable?
| Assignee | ||
Comment 27•14 years ago
|
||
The writing's on the wall on this one: not nearly enough demand to justify leaving this in anyone's review queue after four years of stagnation. I give up. :(
Status: ASSIGNED → RESOLVED
Closed: 14 years ago
Resolution: --- → WONTFIX
| Assignee | ||
Updated•14 years ago
|
Attachment #289196 -
Flags: review?(peterv)
You need to log in
before you can comment on or make changes to this bug.
Description
•