Hi Daniel, On Tue, 19 May 2020, Daniel Wagner wrote: > > I tried to figure out if with the patch the compiler generates different > object code. Looking through the filtered diff between the two versions > I haven't really found any relevant changes. All looks good. > > In case someone wants to look at the diffs: > > https://monom.org/data/qla2xxx/qla2xxx-endianness-annotations.diff > https://monom.org/data/qla2xxx/qla2xxx-endianness-annotations-filtered.diff > I agree. qla2xxx-endianness-annotations.diff seems to be noise. The differences in the __bug_table sections and ql_dbg() call sites are presumably caused by line break changes. Perhaps they can be squelched by inserting blank lines at the appropriate places (for either build). That could probably be automated. Once the comments are stripped from the .s files, we are left with differences like this, @@ -21204,75 +21205,75 @@ .quad .LC146 .quad .LC146 .quad .LC272 + .local __key.63281 + .comm __key.63281,0,1 .local __key.63280 .comm __key.63280,0,1 - .local __key.63279 - .comm __key.63279,0,1 .align 16 - .type __func__.63244, @object - .size __func__.63244, 22 -__func__.63244: + .type __func__.63245, @object + .size __func__.63245, 22 +__func__.63245: .string "qlt_24xx_config_rings" .align 16 .type __func__.61306, @object .size __func__.61306, 26 ... @@ -2571,7 +2571,7 @@ .pushsection __jump_table, "aw" .balign 8 .long 1b - ., .L218 - . # - .quad __UNIQUE_ID_ddebug285.63134+40 + 0 - . #, + .quad __UNIQUE_ID_ddebug285.63135+40 + 0 - . #, .popsection # 0 "" 2 It would be nice to know how these symbols end up with different numbering between builds because it makes a real mess of the diff. I wonder whether the Reproducible Builds project has developed any techniques that could be applied here. https://reproducible-builds.org/
