drivers/fc4/fc.c in git-scsi-misc
davem won't be happy
drivers/fc4/fc.c: In function `fcp_scsi_receive':drivers/fc4/fc.c:397: error: structure has no member named done' drivers/fc4/fc.c:450: error: structure has no member named done' drivers/fc4/fc.c: In function `fcp_scsi_queuecommand':drivers/fc4/fc.c:837: error: structure has no member named done' drivers/fc4/fc.c:838: error: structure has no member named done' drivers/fc4/fc.c:839: error: structure has no member named `done'
Quoted 13 lines · Andrew Morton
davem won't be happy
drivers/fc4/fc.c: In function `fcp_scsi_receive':drivers/fc4/fc.c:397: error: structure has no member nameddone' drivers/fc4/fc.c:450: error: structure has no member nameddone'drivers/fc4/fc.c: In function `fcp_scsi_queuecommand':drivers/fc4/fc.c:837: error: structure has no member nameddone' drivers/fc4/fc.c:838: error: structure has no member nameddone'drivers/fc4/fc.c:839: error: structure has no member named `done' -
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
OK that was a nice lesson in Kconfig and Linux-Makefile for me :)
So I can see that drivers/fc4/Kconfig is only "source"d from arch/sparc64/Kconfig. Hence no one experienced the breakage.
[
If I add to, say, drivrs/Kconfig
+source drivers/fc4/Kconfig
Than everything compiles and I can see the breakage now ]
Do you have a cross-compiling environment to compile all these ARCHs, or you actually have these machines laying around for compilation. (It's the cross-compiling I'm interested in)
Thanks
Boaz
Quoted 17 lines · On Thu, 11 Oct 2007 13:05:32 +0200 Boaz Harrosh <[email protected]> wrote:
Quoted 13 lines · Andrew Morton
davem won't be happy
drivers/fc4/fc.c: In function `fcp_scsi_receive':drivers/fc4/fc.c:397: error: structure has no member nameddone' drivers/fc4/fc.c:450: error: structure has no member nameddone'drivers/fc4/fc.c: In function `fcp_scsi_queuecommand':drivers/fc4/fc.c:837: error: structure has no member nameddone' drivers/fc4/fc.c:838: error: structure has no member nameddone'drivers/fc4/fc.c:839: error: structure has no member named `done' -
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.htmlOK that was a nice lesson in
Kconfigand Linux-Makefile for me :)
Quoted 8 lines
So I can see that
drivers/fc4/Kconfigis only "source"d fromarch/sparc64/Kconfig. Hence no one experienced the breakage.[
If I add to, say, drivrs/Kconfig
+sourcedrivers/fc4/Kconfig
Than everything compiles and I can see the breakage now ]
Well if fc4.c compiles OK on non-sparc64 then perhaps we should enable compilation on non-sparc64. It will increase maintainability and code quality and stuff.
otoh people might end up shipping useless drivers in x86 distros.
Quoted 3 lines
Do you have a cross-compiling environment to compile all these ARCHs, or you actually have these machines laying around for compilation. (It's the cross-compiling I'm interested in)
There are a number of precompiled (by me) cross-compilers at
http://userweb.kernel.org/~akpm/cross-compilers/
From: Andrew Morton <[email protected]>
Date: Thu, 11 Oct 2007 11:20:58 -0700
Quoted 3 lines
Well if
fc4.ccompiles OK on non-sparc64 then perhaps we should enable compilation on non-sparc64. It will increase maintainability and code quality and stuff.
No objections to including drivers/fc4/Kconfig directly from drivers/Kconfig.
It's only included directly from arch/sparc64/Kconfig because long long ago drivers/Kconfig could not be safely included there.
Quoted 7 lines · David Miller
From: Andrew Morton <[email protected]>
Date: Thu, 11 Oct 2007 11:20:58 -0700Quoted 3 lines
Well if
fc4.ccompiles OK on non-sparc64 then perhaps we should enable compilation on non-sparc64. It will increase maintainability and code quality and stuff.
Quoted 6 lines
No objections to including
drivers/fc4/Kconfigdirectly fromdrivers/Kconfig.It's only included directly from
arch/sparc64/Kconfigbecause long long agodrivers/Kconfigcould not be safely included there.
Would you be able to test out Matthew's patch and ack it? Not having fc4 compile is a blocker for the necessary SCSI merge.
James
From: James Bottomley <[email protected]>
Date: Fri, 12 Oct 2007 13:04:33 -0500
Quoted 11 lines · David Miller
Quoted 10 lines
From: Andrew Morton <[email protected]>
Date: Thu, 11 Oct 2007 11:20:58 -0700Quoted 3 lines
Well if
fc4.ccompiles OK on non-sparc64 then perhaps we should enable compilation on non-sparc64. It will increase maintainability and code quality and stuff.No objections to including
drivers/fc4/Kconfigdirectly fromdrivers/Kconfig.
Quoted 6 lines
Quoted 3 lines
It's only included directly from
arch/sparc64/Kconfigbecause long long agodrivers/Kconfigcould not be safely included there.Would you be able to test out Matthew's patch and ack it? Not having fc4 compile is a blocker for the necessary SCSI merge.
Unfortunately it doesn't build. The drivers/scsi/pluto.c part fails with:
drivers/scsi/pluto.c: In function $,1rx(Bpluto_detect$,1ry(B:drivers/scsi/pluto.c:159: error: $,1rx(Bpluto_detect_done$,1ry(B undeclared (first use in this function) drivers/scsi/pluto.c:159: error: (Each undeclared identifier is reported only once drivers/scsi/pluto.c:159: error: for each function it appears in.)
make[1]: *** [drivers/scsi/pluto.o] Error 1
make: *** [drivers/scsi/pluto.o] Error 2
I think it can be fixed by merely removing the assignment on that line, but I'l like Matthew to look at that and generate a new patch if he agrees that is sufficient.
Also, drivers/fc4/fc.c gives a warning:
drivers/fc4/fc.c: In function $,1rx(Bfcp_scsi_abort$,1ry(B:drivers/fc4/fc.c:867: warning: unused variable $,1rx(Bfcmd$,1ry(B drivers/fc4/fc.c: In function $,1rx(B__fcp_scsi_host_reset$,1ry(B:drivers/fc4/fc.c:990: warning: unused variable $,1rx(Bfcmd$,1ry(B
which is fixed thusly:
drivers/fc4/fc.c renamed from drivers/fc4/fc.c~ +0 -2
@@ -864,7 +864,6 @@int fcp_scsi_abort (struct scsi_cmnd *SCpnt) {/* Internal bookkeeping only. Lose 1 cmd_slots slot. */ -fcp_cmnd *fcmd =FCP_CMND (SCpnt);fc_channel *fc =FC_SCMND (SCpnt);/*
@@ -987,7 +986,6 @@static int __fcp_scsi_host_reset (struct scsi_cmnd *SCpnt) {fc_channel *fc =FC_SCMND (SCpnt); -fcp_cmnd *fcmd =FCP_CMND (SCpnt);int i;printk ("FC: host reset\n" );
Quoted 5 lines · David Miller
Unfortunately it doesn't build. The
drivers/scsi/pluto.cpart fails with:
drivers/scsi/pluto.c: In function $,1rxpluto_detect$,1ry:drivers/scsi/pluto.c:159: error: $,1rxpluto_detect_done$,1ry undeclared (first use in this function)
I bet you didn't try this patch on top of scsi-misc. The assignment of pluto_detect_done was removed in an earlier patch, but I neglected to remove the definition of pluto_detect_done.
Looking at that patch again, I may have made the cardinal mistake of believing the comment instead of the code:
I replaced an indirect call to pluto_detect_done with a call to pluto_detect_scsi_done. Whoops. I think the call should simply be deleted (since pluto_detect_done is a no-op).
Quoted 1 line
Also,
drivers/fc4/fc.cgives a warning:
Fix incorporated ... patch against scsi-misc.
drivers/fc4/fc.c +22 -42
@@ -33,6 +33,7 @@#include <linux/slab.h> #include <linux/string.h> #include <linux/init.h> +#include <linux/pci.h> #include <asm/pgtable.h> #include <asm/irq.h>
@@ -378,7 +379,12 @@ void fcp_register(fc_channel *fc, u8 type, int unregister)printk ("FC: %segistering unknown type %02x\n" , unregister ?"Unr" :"R" , type); } -static void fcp_scsi_done (struct scsi_cmnd *SCpnt); +static void fcp_scsi_done (struct scsi_cmnd *SCpnt,unsigned int result) +{ + SCpnt->result = result; +FCD (("Calling scsi_done with %08x\n" , SCpnt->result)) + SCpnt->scsi_done(SCpnt); +}static inline void fcp_scsi_receive (fc_channel *fc,int token,int status,fc_hdr *fch) {
@@ -394,9 +400,6 @@ static inline void fcp_scsi_receive(fc_channel *fc, int token, int status, fc_hd rsp = (fcp_rsp *) (fc->scsi_rsp_pool + fc->rsp_size * token); SCpnt =SC_FCMND (fcmd); -if (SCpnt->done != fcp_scsi_done) -return ; - rsp_status = rsp->fcp_status; FCD(("rsp_status %08x status %08x\n", rsp_status, status))switch (status) {
@@ -442,17 +445,9 @@ static inline void fcp_scsi_receive(fc_channel *fc, int token, int status, fc_hdprintk ("%s: (%d,%d) Received rsp_status 0x%x\n" , fc->name, SCpnt->device->channel, SCpnt->device->id, rsp_status); } - SCpnt->result = (host_status <<16 ) | (rsp_status &0xff ); -#ifdef FCDEBUG -if (host_status || SCpnt->result || rsp_status)printk ("FC: host_status %d, packet status %d\n" , - host_status, SCpnt->result); -#endif - SCpnt->done = fcmd->done; - fcmd->done=NULL ;clear_bit (token, fc->scsi_bitmap); fc->scsi_free++; -FCD (("Calling scsi_done with %08x\n" , SCpnt->result)) - SCpnt->scsi_done(SCpnt); +fcp_scsi_done (SCpnt, (host_status <<16 ) | (rsp_status &0xff )); }void fcp_receive_solicited (fc_channel *fc,int proto,int token,int status,fc_hdr *fch)
@@ -751,13 +746,6 @@ void fcp_release(fc_channel *fcchain, int count) /* count must > 0 */*/ } - -static void fcp_scsi_done (struct scsi_cmnd *SCpnt) -{ -if (FCP_CMND (SCpnt)->done) -FCP_CMND (SCpnt)->done (SCpnt); -} -static int fcp_scsi_queue_it (fc_channel *fc,struct scsi_cmnd *SCpnt,fcp_cmnd *fcmd,int prepare) {
@@ -773,8 +761,7 @@ static int fcp_scsi_queue_it(fc_channel *fc, struct scsi_cmnd *SCpnt,if (fc->encode_addr (SCpnt, cmd->fcp_addr, fc, fcmd)) {/* Invalid channel/id/lun and couldn't map it into fcp_addr */ clear_bit (i, fc->scsi_bitmap); - SCpnt->result = (DID_BAD_TARGET <<16 ); - SCpnt->scsi_done (SCpnt); +fcp_scsi_done (SCpnt, (DID_BAD_TARGET <<16 ));return 0 ; } fc->scsi_free--;
@@ -834,21 +821,20 @@ int fcp_scsi_queuecommand(struct scsi_cmnd *SCpnt,fc_channel *fc =FC_SCMND (SCpnt);FCD (("Entering SCSI queuecommand %p\n" , fcmd)) -if (SCpnt->done != fcp_scsi_done) { - fcmd->done = SCpnt->done; - SCpnt->done = fcp_scsi_done; - SCpnt->scsi_done = done; - fcmd->proto =TYPE_SCSI_FCP ; -if (!fc->scsi_free) { - FCD(("FC: !scsi_free, putting cmd on ML queue\n")) + SCpnt->scsi_done = done; +if (!fc->scsi_free) { + FCD(("FC: !scsi_free, putting cmd on ML queue\n"))#if (FCP_SCSI_USE_NEW_EH_CODE ==0 ) -printk ("fcp_scsi_queue_command: queue full, losing cmd, bad\n" ); +printk ("fcp_scsi_queue_command: queue full, losing cmd, bad\n" );#endif -return 1 ; - } +return 1 ; + } +if (fcmd->proto ==TYPE_SCSI_FCP ) { +return fcp_scsi_queue_it (fc, SCpnt, fcmd,0 ); + }else { + fcmd->proto =TYPE_SCSI_FCP ;return fcp_scsi_queue_it (fc, SCpnt, fcmd,1 ); } -return fcp_scsi_queue_it (fc, SCpnt, fcmd,0 ); }void fcp_queue_empty (fc_channel *fc)
@@ -867,7 +853,6 @@ void fcp_queue_empty(fc_channel *fc)int fcp_scsi_abort (struct scsi_cmnd *SCpnt) {/* Internal bookkeeping only. Lose 1 cmd_slots slot. */ -fcp_cmnd *fcmd =FCP_CMND (SCpnt);fc_channel *fc =FC_SCMND (SCpnt);/*
@@ -890,8 +875,7 @@ int fcp_scsi_abort(struct scsi_cmnd *SCpnt)*/ if (++fc->abort_count < (fc->can_queue >>1 )) { - SCpnt->result =DID_ABORT ; - fcmd->done (SCpnt); +fcp_scsi_done (SCpnt,DID_ABORT );printk ("FC: soft abort\n" );return SUCCESS ; }else {
@@ -949,8 +933,6 @@ int fcp_scsi_dev_reset(struct scsi_cmnd *SCpnt)if (fc->rst_pkt->eh_state ==SCSI_STATE_QUEUED )return FAILED ;/* or SUCCESS. Only these */ } - fc->rst_pkt->done =NULL ; - fc->rst_pkt->eh_state =SCSI_STATE_QUEUED ;init_timer (&fc->rst_pkt->eh_timeout);
@@ -966,7 +948,7 @@ int fcp_scsi_dev_reset(struct scsi_cmnd *SCpnt) fc->rst_pkt->device->host->eh_action = &sem; - fc->rst_pkt->done = fcp_scsi_reset_done; + fc->rst_pkt->scsi_done = fcp_scsi_reset_done;spin_lock_irqsave (SCpnt->device->host->host_lock, flags);fcp_scsi_queue_it (fc, fc->rst_pkt, fcmd,0 );
@@ -993,15 +975,13 @@ int fcp_scsi_dev_reset(struct scsi_cmnd *SCpnt)static int __fcp_scsi_host_reset (struct scsi_cmnd *SCpnt) {fc_channel *fc =FC_SCMND (SCpnt); -fcp_cmnd *fcmd =FCP_CMND (SCpnt);int i;printk ("FC: host reset\n" );for (i=0 ; i < fc->can_queue; i++) {if (fc->cmd_slots[i] && SCpnt->result !=DID_ABORT ) { - SCpnt->result =DID_RESET ; - fcmd->done (SCpnt); +fcp_scsi_done (SCpnt,DID_RESET ); fc->cmd_slots[i] =NULL ; } }
drivers/fc4/fcp_impl.h +0 -1
@@ -39,7 +39,6 @@ struct _fc_channel; typedef struct fcp_cmnd { struct fcp_cmnd *next; struct fcp_cmnd *prev; - void (*done)(struct scsi_cmnd *); unsigned short proto; unsigned short token; unsigned int did;
drivers/scsi/pluto.c +1 -9
@@ -55,14 +55,9 @@ static DECLARE_COMPLETION(fc_detect_complete);static int pluto_encode_addr (Scsi_Cmnd *SCpnt,u16 *addr,fc_channel *fc,fcp_cmnd *fcmd); -static void __init pluto_detect_done (Scsi_Cmnd *SCpnt) -{ -/* Do nothing */ -} -static void __init pluto_detect_scsi_done (Scsi_Cmnd *SCpnt) { - PLND(("Detect done %08lx\n", (long)SCpnt)) + PLND(("Detect done %p\n", SCpnt))if (atomic_dec_and_test (&fcss))complete (&fc_detect_complete); }
@@ -193,9 +188,6 @@ int __init pluto_detect(struct scsi_host_template *tpnt) SCpnt = &(fcs[i].cmd); -/* Let FC mid-level free allocated resources */ -pluto_detect_scsi_done (SCpnt); -if (!SCpnt->result) {struct pluto_inquiry *inq;struct pluto *pluto;
Signature · 4 lines
Intel are signing my paycheques ... these opinions are still mine "Bill, look, we understand that you're interested in selling us this operating system, but compare it to ours. We can't possibly take such a retrograde step."
Quoted 1 line · Andrew Morton
davem won't be happy
I'm not happy either. I think this should fix it, but obviously I have no sparc machines of my own to test it, let alone a pluto.
----
Remove uses of the scsi_cmnd ->done method from the fc4 driver. It was being abused to flag commands that had already been through queuecommand; use the fcmd->proto value for that instead. The fcmd->done pointer now becomes irrelevant. Reuse the fcp_scsi_done name for an entirely different function which provides a handy single place to call ->scsi_done (removing some broken places that used to leak scsi_cmnds on error by calling ->done).
I had to add an include of <linux/pci.h> to get this to compile on non-sparc, and pluto_detect_done() was unused, so remove it too.
drivers/fc4/fc.c +22 -40
@@ -33,6 +33,7 @@#include <linux/slab.h> #include <linux/string.h> #include <linux/init.h> +#include <linux/pci.h> #include <asm/pgtable.h> #include <asm/irq.h>
@@ -378,7 +379,12 @@ void fcp_register(fc_channel *fc, u8 type, int unregister)printk ("FC: %segistering unknown type %02x\n" , unregister ?"Unr" :"R" , type); } -static void fcp_scsi_done (struct scsi_cmnd *SCpnt); +static void fcp_scsi_done (struct scsi_cmnd *SCpnt,unsigned int result) +{ + SCpnt->result = result; +FCD (("Calling scsi_done with %08x\n" , SCpnt->result)) + SCpnt->scsi_done(SCpnt); +}static inline void fcp_scsi_receive (fc_channel *fc,int token,int status,fc_hdr *fch) {
@@ -394,9 +400,6 @@ static inline void fcp_scsi_receive(fc_channel *fc, int token, int status, fc_hd rsp = (fcp_rsp *) (fc->scsi_rsp_pool + fc->rsp_size * token); SCpnt =SC_FCMND (fcmd); -if (SCpnt->done != fcp_scsi_done) -return ; - rsp_status = rsp->fcp_status; FCD(("rsp_status %08x status %08x\n", rsp_status, status))switch (status) {
@@ -442,17 +445,9 @@ static inline void fcp_scsi_receive(fc_channel *fc, int token, int status, fc_hdprintk ("%s: (%d,%d) Received rsp_status 0x%x\n" , fc->name, SCpnt->device->channel, SCpnt->device->id, rsp_status); } - SCpnt->result = (host_status <<16 ) | (rsp_status &0xff ); -#ifdef FCDEBUG -if (host_status || SCpnt->result || rsp_status)printk ("FC: host_status %d, packet status %d\n" , - host_status, SCpnt->result); -#endif - SCpnt->done = fcmd->done; - fcmd->done=NULL ;clear_bit (token, fc->scsi_bitmap); fc->scsi_free++; -FCD (("Calling scsi_done with %08x\n" , SCpnt->result)) - SCpnt->scsi_done(SCpnt); +fcp_scsi_done (SCpnt, (host_status <<16 ) | (rsp_status &0xff )); }void fcp_receive_solicited (fc_channel *fc,int proto,int token,int status,fc_hdr *fch)
@@ -751,13 +746,6 @@ void fcp_release(fc_channel *fcchain, int count) /* count must > 0 */*/ } - -static void fcp_scsi_done (struct scsi_cmnd *SCpnt) -{ -if (FCP_CMND (SCpnt)->done) -FCP_CMND (SCpnt)->done (SCpnt); -} -static int fcp_scsi_queue_it (fc_channel *fc,struct scsi_cmnd *SCpnt,fcp_cmnd *fcmd,int prepare) {
@@ -773,8 +761,7 @@ static int fcp_scsi_queue_it(fc_channel *fc, struct scsi_cmnd *SCpnt,if (fc->encode_addr (SCpnt, cmd->fcp_addr, fc, fcmd)) {/* Invalid channel/id/lun and couldn't map it into fcp_addr */ clear_bit (i, fc->scsi_bitmap); - SCpnt->result = (DID_BAD_TARGET <<16 ); - SCpnt->scsi_done (SCpnt); +fcp_scsi_done (SCpnt, (DID_BAD_TARGET <<16 ));return 0 ; } fc->scsi_free--;
@@ -834,21 +821,20 @@ int fcp_scsi_queuecommand(struct scsi_cmnd *SCpnt,fc_channel *fc =FC_SCMND (SCpnt);FCD (("Entering SCSI queuecommand %p\n" , fcmd)) -if (SCpnt->done != fcp_scsi_done) { - fcmd->done = SCpnt->done; - SCpnt->done = fcp_scsi_done; - SCpnt->scsi_done = done; - fcmd->proto =TYPE_SCSI_FCP ; -if (!fc->scsi_free) { - FCD(("FC: !scsi_free, putting cmd on ML queue\n")) + SCpnt->scsi_done = done; +if (!fc->scsi_free) { + FCD(("FC: !scsi_free, putting cmd on ML queue\n"))#if (FCP_SCSI_USE_NEW_EH_CODE ==0 ) -printk ("fcp_scsi_queue_command: queue full, losing cmd, bad\n" ); +printk ("fcp_scsi_queue_command: queue full, losing cmd, bad\n" );#endif -return 1 ; - } +return 1 ; + } +if (fcmd->proto ==TYPE_SCSI_FCP ) { +return fcp_scsi_queue_it (fc, SCpnt, fcmd,0 ); + }else { + fcmd->proto =TYPE_SCSI_FCP ;return fcp_scsi_queue_it (fc, SCpnt, fcmd,1 ); } -return fcp_scsi_queue_it (fc, SCpnt, fcmd,0 ); }void fcp_queue_empty (fc_channel *fc)
@@ -890,8 +876,7 @@ int fcp_scsi_abort(struct scsi_cmnd *SCpnt)*/ if (++fc->abort_count < (fc->can_queue >>1 )) { - SCpnt->result =DID_ABORT ; - fcmd->done (SCpnt); +fcp_scsi_done (SCpnt,DID_ABORT );printk ("FC: soft abort\n" );return SUCCESS ; }else {
@@ -949,8 +934,6 @@ int fcp_scsi_dev_reset(struct scsi_cmnd *SCpnt)if (fc->rst_pkt->eh_state ==SCSI_STATE_QUEUED )return FAILED ;/* or SUCCESS. Only these */ } - fc->rst_pkt->done =NULL ; - fc->rst_pkt->eh_state =SCSI_STATE_QUEUED ;init_timer (&fc->rst_pkt->eh_timeout);
@@ -966,7 +949,7 @@ int fcp_scsi_dev_reset(struct scsi_cmnd *SCpnt) fc->rst_pkt->device->host->eh_action = &sem; - fc->rst_pkt->done = fcp_scsi_reset_done; + fc->rst_pkt->scsi_done = fcp_scsi_reset_done;spin_lock_irqsave (SCpnt->device->host->host_lock, flags);fcp_scsi_queue_it (fc, fc->rst_pkt, fcmd,0 );
@@ -1000,8 +983,7 @@ static int __fcp_scsi_host_reset(struct scsi_cmnd *SCpnt)for (i=0 ; i < fc->can_queue; i++) {if (fc->cmd_slots[i] && SCpnt->result !=DID_ABORT ) { - SCpnt->result =DID_RESET ; - fcmd->done (SCpnt); +fcp_scsi_done (SCpnt,DID_RESET ); fc->cmd_slots[i] =NULL ; } }
drivers/fc4/fcp_impl.h +0 -1
@@ -39,7 +39,6 @@ struct _fc_channel; typedef struct fcp_cmnd { struct fcp_cmnd *next; struct fcp_cmnd *prev; - void (*done)(struct scsi_cmnd *); unsigned short proto; unsigned short token; unsigned int did;
drivers/scsi/pluto.c +0 -5
@@ -55,11 +55,6 @@ static DECLARE_COMPLETION(fc_detect_complete);static int pluto_encode_addr (Scsi_Cmnd *SCpnt,u16 *addr,fc_channel *fc,fcp_cmnd *fcmd); -static void __init pluto_detect_done (Scsi_Cmnd *SCpnt) -{ -/* Do nothing */ -} -static void __init pluto_detect_scsi_done (Scsi_Cmnd *SCpnt) {PLND (("Detect done %08lx\n" , (long )SCpnt))
Signature · 4 lines
Intel are signing my paycheques ... these opinions are still mine "Bill, look, we understand that you're interested in selling us this operating system, but compare it to ours. We can't possibly take such a retrograde step."