From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gerald Pfeifer To: Paul Derbyshire Cc: egcs@egcs.cygnus.com Subject: Re: as problem on sgi Date: Sat, 20 Feb 1999 16:48:00 -0000 Message-id: In-reply-to: < 3.0.6.32.19990220115352.0083a100@pop.globalserve.net > References: <3.0.6.32.19990220115352.0083a100@pop.globalserve.net> X-SW-Source: 1999-02/msg01006.html On Sat, 20 Feb 1999, Paul Derbyshire wrote: >> An introductory answer for this is found on >> http://egcs.cygnus.com/faq.html#squangle > It didn't look to me like the C++ symbol was just too long for the > assembler; it looked to me like the symbol was ridiculously long and > repetitive. It seemed to be the same few components repeated dozens of > times, which looks like a symbol generation error... should a symbol really > be output with that many repetitions of "string_char_traits" inside of it? That's exactly the point of -fsquangle, which is mentioned in the FAQ entry above. Note that symbols of these kind mostly happen with template heavy libraries like STL & Co, which are rather "new". -fsquangle makes the generated symbols significantly shorter, but it breaks link compatiblity and thus will be only enabled together with the new C++ ABI. (One does not want to break link compatibility too often...) Gerald -- Gerald Pfeifer (Jerry) Vienna University of Technology pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/ From mboxrd@z Thu Jan 1 00:00:00 1970 From: Gerald Pfeifer To: Paul Derbyshire Cc: egcs@egcs.cygnus.com Subject: Re: as problem on sgi Date: Sun, 28 Feb 1999 22:53:00 -0000 Message-ID: References: <3.0.6.32.19990220115352.0083a100@pop.globalserve.net> X-SW-Source: 1999-02n/msg01006.html Message-ID: <19990228225300.5NK8aojFVgrWkYinxYABwqFzThqX7682rKzQoP9E8ac@z> On Sat, 20 Feb 1999, Paul Derbyshire wrote: >> An introductory answer for this is found on >> http://egcs.cygnus.com/faq.html#squangle > It didn't look to me like the C++ symbol was just too long for the > assembler; it looked to me like the symbol was ridiculously long and > repetitive. It seemed to be the same few components repeated dozens of > times, which looks like a symbol generation error... should a symbol really > be output with that many repetitions of "string_char_traits" inside of it? That's exactly the point of -fsquangle, which is mentioned in the FAQ entry above. Note that symbols of these kind mostly happen with template heavy libraries like STL & Co, which are rather "new". -fsquangle makes the generated symbols significantly shorter, but it breaks link compatiblity and thus will be only enabled together with the new C++ ABI. (One does not want to break link compatibility too often...) Gerald -- Gerald Pfeifer (Jerry) Vienna University of Technology pfeifer@dbai.tuwien.ac.at http://www.dbai.tuwien.ac.at/~pfeifer/