https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92606
--- Comment #7 from Georg-Johann Lay <gjl at gcc dot gnu.org> --- (In reply to Richard Biener from comment #6) > or whether the backend "forgets" to set DECL_SECTION_NAME or the like. That caused problems (don't remember which ones exactly, maybe to make -fdata-sections work). The section is set in TARGET_ASM_SELECT_SECTION so the middle-end knows the section (in principle). But presumably the middle-end concept is broken, because telling what section belongs to a decl is not enough to convince the middle-end that the section belongs to the decl... It's gcc after all. (Very) old versions did set DECL_SECTION_NAME in insert_attributes, but that's no more the case. Presumably to fix one of the 100s of FIXMEs in the avr backend. > Similarly address-spaces need to be honored (I bet avr would have one -- is > PROGMEM actually an address-space implementation-wise?) No, progmem is just an attribute and dates back to the times when there was no address-spaces in gcc. It's still supposed to be supported because old code should still work, and because g++ will never have address-spaces. > I see there's a 'progmem' target attribute but its effects might be hidden > completely from the middle-end. That's why I opened PR92932. As mentioned in comment #3, the bug also appears for address-space __flash.
