diff mbox series

RTEMS: Improve GCC specification

Message ID 20200408174721.14408-1-sebastian.huber@embedded-brains.de
State New
Headers show
Series RTEMS: Improve GCC specification | expand

Commit Message

Sebastian Huber April 8, 2020, 5:47 p.m. UTC 
Add a start/end file specification if the -qrtems option is present.
Allow targets to customize it.

Support the standard -nodefaultlibs option.

gcc/

	* config/rtems.h (RTEMS_STARTFILE_SPEC): Define if undefined.
	(RTEMS_ENDFILE_SPEC): Likewise.
	(STARTFILE_SPEC): Update comment.  Add RTEMS_STARTFILE_SPEC.
	(ENDFILE_SPEC): Add RTEMS_ENDFILE_SPEC.
	(LIB_SPECS): Support -nodefaultlibs option.
	* config/or1k/rtems.h (RTEMS_STARTFILE_SPEC): Define.
	(RTEMS_ENDFILE_SPEC): Likewise.
	* config/rs6000/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
	(RTEMS_ENDFILE_SPEC): Likewise.
	* config/v850/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
	(RTEMS_ENDFILE_SPEC): Likewise.
---
 gcc/config/or1k/rtems.h   |  3 +++
 gcc/config/rs6000/rtems.h |  3 +++
 gcc/config/rtems.h        | 26 ++++++++++++++++++--------
 gcc/config/v850/rtems.h   |  3 +++
 4 files changed, 27 insertions(+), 8 deletions(-)

Comments

Li, Pan2 via Gcc-patches April 8, 2020, 8:24 p.m. UTC | #1 
On Wed, 2020-04-08 at 19:47 +0200, Sebastian Huber wrote:
> Add a start/end file specification if the -qrtems option is present.
> Allow targets to customize it.
> 
> Support the standard -nodefaultlibs option.
> 
> gcc/
> 
> 	* config/rtems.h (RTEMS_STARTFILE_SPEC): Define if undefined.
> 	(RTEMS_ENDFILE_SPEC): Likewise.
> 	(STARTFILE_SPEC): Update comment.  Add RTEMS_STARTFILE_SPEC.
> 	(ENDFILE_SPEC): Add RTEMS_ENDFILE_SPEC.
> 	(LIB_SPECS): Support -nodefaultlibs option.
> 	* config/or1k/rtems.h (RTEMS_STARTFILE_SPEC): Define.
> 	(RTEMS_ENDFILE_SPEC): Likewise.
> 	* config/rs6000/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
> 	(RTEMS_ENDFILE_SPEC): Likewise.
> 	* config/v850/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
> 	(RTEMS_ENDFILE_SPEC): Likewise.
How important is it to get this into gcc-10?  We're well into stage4 at this
point and while we do have leeway with a patch like this we do want to be
selective about what we accept.

If it's important, then OK, otherwise please defer it to gcc-11, where it's pre-
approved.

Jeff
Sebastian Huber April 14, 2020, 3:26 p.m. UTC | #2 
Hello Jeff,

On 08/04/2020 22:24, Jeff Law wrote:

> On Wed, 2020-04-08 at 19:47 +0200, Sebastian Huber wrote:
>> Add a start/end file specification if the -qrtems option is present.
>> Allow targets to customize it.
>>
>> Support the standard -nodefaultlibs option.
>>
>> gcc/
>>
>> 	* config/rtems.h (RTEMS_STARTFILE_SPEC): Define if undefined.
>> 	(RTEMS_ENDFILE_SPEC): Likewise.
>> 	(STARTFILE_SPEC): Update comment.  Add RTEMS_STARTFILE_SPEC.
>> 	(ENDFILE_SPEC): Add RTEMS_ENDFILE_SPEC.
>> 	(LIB_SPECS): Support -nodefaultlibs option.
>> 	* config/or1k/rtems.h (RTEMS_STARTFILE_SPEC): Define.
>> 	(RTEMS_ENDFILE_SPEC): Likewise.
>> 	* config/rs6000/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
>> 	(RTEMS_ENDFILE_SPEC): Likewise.
>> 	* config/v850/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
>> 	(RTEMS_ENDFILE_SPEC): Likewise.
> How important is it to get this into gcc-10?  We're well into stage4 at this
> point and while we do have leeway with a patch like this we do want to be
> selective about what we accept.
>
> If it's important, then OK, otherwise please defer it to gcc-11, where it's pre-
> approved.
would it be all right to commit it to the trunk once the GCC 11 
development starts and then back port it for GCC 10.2 after a while?
Li, Pan2 via Gcc-patches April 15, 2020, 3:25 p.m. UTC | #3 
On Tue, 2020-04-14 at 17:26 +0200, Sebastian Huber wrote:
> Hello Jeff,
> 
> On 08/04/2020 22:24, Jeff Law wrote:
> 
> > On Wed, 2020-04-08 at 19:47 +0200, Sebastian Huber wrote:
> > > Add a start/end file specification if the -qrtems option is present.
> > > Allow targets to customize it.
> > > 
> > > Support the standard -nodefaultlibs option.
> > > 
> > > gcc/
> > > 
> > > 	* config/rtems.h (RTEMS_STARTFILE_SPEC): Define if undefined.
> > > 	(RTEMS_ENDFILE_SPEC): Likewise.
> > > 	(STARTFILE_SPEC): Update comment.  Add RTEMS_STARTFILE_SPEC.
> > > 	(ENDFILE_SPEC): Add RTEMS_ENDFILE_SPEC.
> > > 	(LIB_SPECS): Support -nodefaultlibs option.
> > > 	* config/or1k/rtems.h (RTEMS_STARTFILE_SPEC): Define.
> > > 	(RTEMS_ENDFILE_SPEC): Likewise.
> > > 	* config/rs6000/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
> > > 	(RTEMS_ENDFILE_SPEC): Likewise.
> > > 	* config/v850/rtems.h (RTEMS_STARTFILE_SPEC): Likewise.
> > > 	(RTEMS_ENDFILE_SPEC): Likewise.
> > How important is it to get this into gcc-10?  We're well into stage4 at this
> > point and while we do have leeway with a patch like this we do want to be
> > selective about what we accept.
> > 
> > If it's important, then OK, otherwise please defer it to gcc-11, where it's
> > pre-
> > approved.
> would it be all right to commit it to the trunk once the GCC 11 
> development starts and then back port it for GCC 10.2 after a while?
That works for me.  It's not technically a regression, but I think some leeway is
reasonable.

jeff
diff mbox series

Patch

diff --git a/gcc/config/or1k/rtems.h b/gcc/config/or1k/rtems.h
index 94731435f44..d4b8fef3dc7 100644
--- a/gcc/config/or1k/rtems.h
+++ b/gcc/config/or1k/rtems.h
@@ -28,3 +28,6 @@ 
 	builtin_assert ("system=rtems");	\
     }						\
   while (0)
+
+#define RTEMS_STARTFILE_SPEC "crtbegin%O%s"
+#define RTEMS_ENDFILE_SPEC "crtend%O%s"
diff --git a/gcc/config/rs6000/rtems.h b/gcc/config/rs6000/rtems.h
index 6ce05bfc6d6..40db6606f17 100644
--- a/gcc/config/rs6000/rtems.h
+++ b/gcc/config/rs6000/rtems.h
@@ -292,3 +292,6 @@ 
   "%{mads|myellowknife|mmvme|msim:%G %L %G;" \
   "!mcall-*|mcall-linux:" GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC ";" \
   ":%G %L %G}"
+
+#define RTEMS_STARTFILE_SPEC "ecrti%O%s rtems_crti%O%s crtbegin%O%s"
+#define RTEMS_ENDFILE_SPEC "crtend%O%s rtems_crtn%O%s ecrtn%O%s"
diff --git a/gcc/config/rtems.h b/gcc/config/rtems.h
index 55d568f83b5..30ad6258949 100644
--- a/gcc/config/rtems.h
+++ b/gcc/config/rtems.h
@@ -22,15 +22,25 @@ 
    see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
    <http://www.gnu.org/licenses/>.  */
 
+#ifndef RTEMS_STARTFILE_SPEC
+#define RTEMS_STARTFILE_SPEC "crti%O%s crtbegin%O%s"
+#endif
+
+#ifndef RTEMS_ENDFILE_SPEC
+#define RTEMS_ENDFILE_SPEC "crtend%O%s crtn%O%s"
+#endif
+
 /*
- * Dummy start/end specification to let linker work as
- * needed by autoconf scripts using this compiler.
+ * The crt0.o is a dummy start file to let the linker work as needed by
+ * autoconf scripts using this compiler.
  */
 #undef STARTFILE_SPEC
-#define STARTFILE_SPEC "crt0.o%s"
+#define STARTFILE_SPEC "%{!qrtems:crt0%O%s} " \
+"%{qrtems:%{!nostdlib:%{!nostartfiles:" RTEMS_STARTFILE_SPEC "}}}"
 
 #undef ENDFILE_SPEC
-#define ENDFILE_SPEC   ""
+#define ENDFILE_SPEC \
+"%{qrtems:%{!nostdlib:%{!nostartfiles:" RTEMS_ENDFILE_SPEC "}}}"
 
 /*
  * Some targets do not set up LIB_SPECS, override it, here.
@@ -38,10 +48,10 @@ 
 #define STD_LIB_SPEC "%{!shared:%{g*:-lg} %{!p:%{!pg:-lc}}%{p:-lc_p}%{pg:-lc_p}}"
 
 #undef LIB_SPEC
-#define LIB_SPEC "%{!qrtems: " STD_LIB_SPEC "} " \
-"%{!nostdlib: %{qrtems: --start-group \
- -lrtemsbsp -lrtemscpu \
- -latomic -lc -lgcc --end-group %{!qnolinkcmds: -T linkcmds%s}}}"
+#define LIB_SPEC "%{!qrtems:" STD_LIB_SPEC "} " \
+"%{qrtems:%{!nostdlib:%{!nodefaultlibs:" \
+"--start-group -lrtemsbsp -lrtemscpu -latomic -lc -lgcc --end-group} " \
+"%{!qnolinkcmds:-T linkcmds%s}}}"
 
 #define TARGET_POSIX_IO
 
diff --git a/gcc/config/v850/rtems.h b/gcc/config/v850/rtems.h
index 3f7d39a2cd1..e261ea3b417 100644
--- a/gcc/config/v850/rtems.h
+++ b/gcc/config/v850/rtems.h
@@ -39,3 +39,6 @@ 
 %{!mv850es:%{!mv850e1:%{mv*:-mv%*}} \
 %{m8byte-align:-m8byte-align} \
 %{mgcc-abi:-mgcc-abi}}"
+
+#define RTEMS_STARTFILE_SPEC ""
+#define RTEMS_ENDFILE_SPEC ""