Message ID | mvm36gletqb.fsf@suse.de |
---|---|
State | New |
Headers | show |
Series | sysdeps/riscv/start.S: rename .Lload_gp to load_gp (bug 24376) | expand |
* Andreas Schwab: > Renaming the symbol keeps the label in the symbol table as a normal local > label, so that `objdump -d' can pick it up as an anchor. > > [BZ #24376] > * sysdeps/riscv/start.S (load_gp): Rename from .Lload_gp, all uses > changed. Perhaps use startup_load_gp instead, so that it's less ambiguous in the debugging information?
On Sep 24 2019, Florian Weimer <fw@deneb.enyo.de> wrote: > * Andreas Schwab: > >> Renaming the symbol keeps the label in the symbol table as a normal local >> label, so that `objdump -d' can pick it up as an anchor. >> >> [BZ #24376] >> * sysdeps/riscv/start.S (load_gp): Rename from .Lload_gp, all uses >> changed. > > Perhaps use startup_load_gp instead, so that it's less ambiguous in > the debugging information? It's a local label, so no conflict possible. Andreas.
On Tue, Sep 24, 2019 at 9:23 AM Andreas Schwab <schwab@suse.de> wrote: > On Sep 24 2019, Florian Weimer <fw@deneb.enyo.de> wrote: > > * Andreas Schwab: > >> Renaming the symbol keeps the label in the symbol table as a normal local > >> label, so that `objdump -d' can pick it up as an anchor. > > > > Perhaps use startup_load_gp instead, so that it's less ambiguous in > > the debugging information? > > It's a local label, so no conflict possible. That's not what he was saying. zw
On Sep 24 2019, Zack Weinberg <zackw@panix.com> wrote: > On Tue, Sep 24, 2019 at 9:23 AM Andreas Schwab <schwab@suse.de> wrote: >> On Sep 24 2019, Florian Weimer <fw@deneb.enyo.de> wrote: >> > * Andreas Schwab: >> >> Renaming the symbol keeps the label in the symbol table as a normal local >> >> label, so that `objdump -d' can pick it up as an anchor. >> > >> > Perhaps use startup_load_gp instead, so that it's less ambiguous in >> > the debugging information? >> >> It's a local label, so no conflict possible. > > That's not what he was saying. That's because I said that. Andreas.
On Tue, Sep 24, 2019 at 9:32 AM Andreas Schwab <schwab@suse.de> wrote: > On Sep 24 2019, Zack Weinberg <zackw@panix.com> wrote: > > On Tue, Sep 24, 2019 at 9:23 AM Andreas Schwab <schwab@suse.de> wrote: > >> On Sep 24 2019, Florian Weimer <fw@deneb.enyo.de> wrote: > >> > > >> > Perhaps use startup_load_gp instead, so that it's less ambiguous in > >> > the debugging information? > >> > >> It's a local label, so no conflict possible. > > > > That's not what he was saying. > > That's because I said that. Yes, the issue is that your response does not address his concern. zw
diff --git a/sysdeps/riscv/start.S b/sysdeps/riscv/start.S index bcbce58ffe..dee984131c 100644 --- a/sysdeps/riscv/start.S +++ b/sysdeps/riscv/start.S @@ -47,7 +47,7 @@ ENTRY (ENTRY_POINT) .cfi_label to force starting the FDE. */ .cfi_label .Ldummy cfi_undefined (ra) - call .Lload_gp + call load_gp mv a5, a0 /* rtld_fini. */ /* main may be in a shared library. */ la a0, main @@ -68,7 +68,7 @@ END (ENTRY_POINT) needs to be initialized before calling __libc_start_main in that case. So we redundantly initialize it at the beginning of _start. */ -.Lload_gp: +load_gp: .option push .option norelax lla gp, __global_pointer$ @@ -76,7 +76,7 @@ END (ENTRY_POINT) ret .section .preinit_array,"aw" - .dc.a .Lload_gp + .dc.a load_gp /* Define a symbol for the first piece of initialized data. */ .data