[9fans] broken contrib source
Upon trying to build /sys/src/cmd for the arm today (beagleboards came in), I found that mk was dying due to bad source installed by contrib(1) packages. mk -k install was able to get me through most of it, but the three worst offenders didn't even have mkfiles in their top levels; since I didn't really care about having them, I deleted the directories and continued. Those three offenders: 1. contrib, which seems to have only one file in it. 2. tcl 3. vim And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. Please, those of you who make contrib packages, don't dump source into /sys/src without first checking that it can actually build! It's not too difficult to do a mk install in /sys/src/cmd after you put the source in there. Maybe there should be a contrib/verify script which goes through /sys/src and makes sure everything can still build normally? John -- Object-oriented design is the roman numerals of computing -- Rob Pike
Re: [9fans] broken contrib source
Which machine is this on? I don't quite understand what you're saying. Aren't the contrib packages you're talking about something outside the main distribution? On Thu, Aug 20, 2009 at 12:02 AM, John Florenslawmas...@gmail.com wrote: Upon trying to build /sys/src/cmd for the arm today (beagleboards came in), I found that mk was dying due to bad source installed by contrib(1) packages. mk -k install was able to get me through most of it, but the three worst offenders didn't even have mkfiles in their top levels; since I didn't really care about having them, I deleted the directories and continued. Those three offenders: 1. contrib, which seems to have only one file in it. 2. tcl 3. vim And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. Please, those of you who make contrib packages, don't dump source into /sys/src without first checking that it can actually build! It's not too difficult to do a mk install in /sys/src/cmd after you put the source in there. Maybe there should be a contrib/verify script which goes through /sys/src and makes sure everything can still build normally? John -- Object-oriented design is the roman numerals of computing -- Rob Pike
Re: [9fans] broken contrib source
On Wed Aug 19 18:05:36 EDT 2009, slawmas...@gmail.com wrote: And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. nupas compiles with [78q]c. i hadn't considered compiling it for arm. sorry. however, this appears to be a compiler limitation. it appears not handle ?: with vlongs. neither cache.c, imap.c nor mdir.c compile. all with diag. cache.c:245 unknown vlong LIST i believe a machcap function needs to be written. - erik
Re: [9fans] broken contrib source
Aaaah. now I get it. I found it in fgb's tree. Has any announcing config(1) on the list? On Thu, Aug 20, 2009 at 12:20 AM, Noah Evansnoah.ev...@gmail.com wrote: Which machine is this on? I don't quite understand what you're saying. Aren't the contrib packages you're talking about something outside the main distribution? On Thu, Aug 20, 2009 at 12:02 AM, John Florenslawmas...@gmail.com wrote: Upon trying to build /sys/src/cmd for the arm today (beagleboards came in), I found that mk was dying due to bad source installed by contrib(1) packages. mk -k install was able to get me through most of it, but the three worst offenders didn't even have mkfiles in their top levels; since I didn't really care about having them, I deleted the directories and continued. Those three offenders: 1. contrib, which seems to have only one file in it. 2. tcl 3. vim And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. Please, those of you who make contrib packages, don't dump source into /sys/src without first checking that it can actually build! It's not too difficult to do a mk install in /sys/src/cmd after you put the source in there. Maybe there should be a contrib/verify script which goes through /sys/src and makes sure everything can still build normally? John -- Object-oriented design is the roman numerals of computing -- Rob Pike
Re: [9fans] broken contrib source
Please, those of you who make contrib packages, don't dump source into /sys/src without first checking that it can actually build! It's not too difficult to do a mk install in /sys/src/cmd after you put the source in there. Maybe there should be a contrib/verify script which goes through /sys/src and makes sure everything can still build normally? relax, no need to be a prick -- Federico G. Benavento
Re: [9fans] broken contrib source
On Wed Aug 19 18:28:05 EDT 2009, quans...@coraid.com wrote: On Wed Aug 19 18:05:36 EDT 2009, slawmas...@gmail.com wrote: And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. nupas compiles with [78q]c. i hadn't considered compiling it for arm. sorry. however, this appears to be a compiler limitation. it appears not handle ?: with vlongs. neither cache.c, imap.c nor mdir.c compile. all with diag. cache.c:245 unknown vlong LIST i believe a machcap function needs to be written. it compiles now, but i can't check it: ; diffy -c /sys/src/cmd/cc/com64.c /n/dump/2009/0819/sys/src/cmd/cc/com64.c:274,279 - /sys/src/cmd/cc/com64.c:274,280 case ORETURN: case OAS: case OIND: + case OLIST: return 1; case OADD: a = nodaddv; - erik
Re: [9fans] broken contrib source
Fixed contrib and tcl. the latter had a :? with vlongs that needed a change. thanks On Wed, Aug 19, 2009 at 7:41 PM, erik quanstromquans...@coraid.com wrote: On Wed Aug 19 18:28:05 EDT 2009, quans...@coraid.com wrote: On Wed Aug 19 18:05:36 EDT 2009, slawmas...@gmail.com wrote: And, while it didn't bring mk -k to a grinding halt, I did notice that, with nupas installed, cd /sys/src/cmd/upas; objtype=arm mk install chokes on /sys/src/cmd/upas/imap4.c/imap4.c. nupas compiles with [78q]c. i hadn't considered compiling it for arm. sorry. however, this appears to be a compiler limitation. it appears not handle ?: with vlongs. neither cache.c, imap.c nor mdir.c compile. all with diag. cache.c:245 unknown vlong LIST i believe a machcap function needs to be written. it compiles now, but i can't check it: ; diffy -c /sys/src/cmd/cc/com64.c /n/dump/2009/0819/sys/src/cmd/cc/com64.c:274,279 - /sys/src/cmd/cc/com64.c:274,280 case ORETURN: case OAS: case OIND: + case OLIST: return 1; case OADD: a = nodaddv; - erik -- Federico G. Benavento
