Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
On 17.02.2012 00:20, Kohei Yoshida wrote: As far as I know those STL methods that return its element are guaranteed to return a reference to the stored instance, not its copy. Ah, great, I was not sure of that. On 17.02.2012 01:01, Michael Meeks wrote: I guess :-) I imagine the real problem here is that any kind of operation on the stl container could then realloc the underlying memory in some way, busting the reference / pointer you handed out elsewhere; so it looks only transiently safe to me ;-) I inspected the usages of that function - they look safe. However, the details are elusive for the users... Thanks for the answers! Best Regards, Ivan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
On Thu, Feb 16, 2012 at 4:01 PM, Michael Meeks wrote: > > On Thu, 2012-02-16 at 15:20 -0500, Kohei Yoshida wrote: >> > Is it a safe code? >> > >> > pForbiddenCharacters = &maMap[ nLanguage ]; >> > return pForbiddenCharacters; > ... >> So, in theory, taking the address of a returned object which itself is a >> reference to the instance stored in the container is very much safe. > > I guess :-) I imagine the real problem here is that any kind of > operation on the stl container could then realloc the underlying memory > in some way, busting the reference / pointer you handed out elsewhere; > so it looks only transiently safe to me ;-) Of course. If the content of the container changes all bets are off. But I thought that was a given in this context (the same way all associated iterators are invalid once the state of the container changes). My understanding of the question was, does the address of the object point to the real object inside the container or a temporary object? And my answer was it points to the real object inside the container. I didn't mean to insinuate anything else. Sorry if I gave such impression. Kohei ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
On Thu, 2012-02-16 at 15:20 -0500, Kohei Yoshida wrote: > > Is it a safe code? > > > > pForbiddenCharacters = &maMap[ nLanguage ]; > > return pForbiddenCharacters; ... > So, in theory, taking the address of a returned object which itself is a > reference to the instance stored in the container is very much safe. I guess :-) I imagine the real problem here is that any kind of operation on the stl container could then realloc the underlying memory in some way, busting the reference / pointer you handed out elsewhere; so it looks only transiently safe to me ;-) Or perhaps I'm missing something, ATB, Michael. -- michael.me...@suse.com <><, Pseudo Engineer, itinerant idiot ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
On Thu, 2012-02-16 at 10:48 +0400, Ivan Timofeev wrote: > Whoa, I seem to be too naive for patch-reviewing... > > Is it a safe code? > > pForbiddenCharacters = &maMap[ nLanguage ]; > return pForbiddenCharacters; Yeah I don't see anything wrong with it. > Will pForbiddenCharacters point to the deleted object, after the return? Nope. As far as I know those STL methods that return its element are guaranteed to return a reference to the stored instance, not its copy. So, in theory, taking the address of a returned object which itself is a reference to the instance stored in the container is very much safe. The only exception is std::vector which lumps its internal boolean values into bitfield, but by doing it it can no longer meet the requirement of STL container which must return reference to its elements. So, by that definition std::vector is not a STL container, strictly speaking. For the curious, refer to Item 18 of Effective STL by Scott Meyers. Kohei -- Kohei Yoshida, LibreOffice hacker, Calc ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
Whoa, I seem to be too naive for patch-reviewing... Is it a safe code? pForbiddenCharacters = &maMap[ nLanguage ]; return pForbiddenCharacters; Will pForbiddenCharacters point to the deleted object, after the return? Terribly sorry for such a question... Regards, Ivan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
Re: [PUSHED][PATCH] convert table.hxx use in editeng/inc/editeng/forbiddencharacterstable.hxx
Hi Noel, On 15.02.2012 19:01, Noel Grandin wrote: Updated patch implementing Ivan's suggestion. Passes make and make check. Convert tools/table.hxx usage in editeng/inc/editeng/forbiddencharacterstable.hxx to std::map Looks good to me, pushed (without GetCharInfo - it was unused): http://cgit.freedesktop.org/libreoffice/core/commit/?id=9d4b0c25a598a53601e2bd337443728f17f8296a and tweaked: http://cgit.freedesktop.org/libreoffice/core/commit/?id=843eafc765a3d1d0ea4c9a89855c73e81784aa8b (I think, having 'Map' as a type alias is enough... :) Thanks for your work! Regards, Ivan ___ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
