#tyndur · 2010-02-12

536 Nachrichten · Seite 1 von 3 · Original-CSV herunterladen

  1. taljethgn8
  2. XanClicGood idea
  3. XanClicgn8
  4. kwolfMoin
  5. pc[away]Patch coming your way within 30 minutes, kwolf
  6. kwolfHi pcmattman
  7. kwolfSounds good
  8. pcmattmanRight after that I'll see what I can and can't do in this submodule
  9. kwolfYup, we should find a solution that keeps Pedigree compilable most of the time
  10. pcmattmanAnd sent
  11. pcmattmanAh, should've used [PATCH v2]
  12. pcmattmanAw, it also removed my explanation paragraph... odd
  13. pcmattmanHeh, next time I'll remember to *save* the modified .patch file
  14. kwolfMy git doesn't remove patch files ;-)
  15. pcmattmanNo point emailing it again, the patch will just have to document the change :)
  16. kwolfI think it's fine as it is
  17. kwolfEven thought the subject is rather long ;-)
  18. pcmattmanHehe
  19. pcmattmanNow, as for checking out that remote branch... how do I go about that?
  20. kwolfHm, assuming that the submodule is just a normal repo...
  21. kwolfSo git checkout origin/usb in the CDI directory?
  22. kwolfYou can do a git pull there, right?
  23. kwolfSo it should be more or less a normal repo
  24. pcmattmanNote: moving to 'origin/usb' which isn't a local branch
  25. pcmattman$ git branch -v
  26. kwolf"Note", not "Error" ;-)
  27. pcmattman* (no branch) 8b00998 usb: A bit faster
  28. pcmattman master 610dc3e [ahead 1] Change CDI headers to use angle brackets for other CDI headers
  29. pcmattmanIt certainly seems to work right
  30. pcmattmanIt just has no idea what branch I just moved to
  31. kwolfRight, the question is if you could commit this as the submodule revision
  32. kwolfThe branch you moved to is origin/usb, obviously ;)
  33. kwolfYou can't move the head of this branch because its a remote branch, so it detaches the head. It's like when you checkout a specific commit
  34. pcmattmanHmm, well
  35. pcmattmanThere's a full .git in the submodule root
  36. kwolfYes, I think it's a real repo
  37. kwolfWhose head is referenced by pedigree.git
  38. pcmattmanHmm, perhaps
  39. pcmattmanWhat's cdi.git's master's current HEAD commit?
  40. kwolfDoes git diff show a changed submodule head?
  41. kwolforigin/master ;-)
  42. pcmattmanI mean, the hash ;)
  43. kwolfgitweb says 2474abcf
  44. pcmattmanOkay
  45. pcmattmanI made a commit into the non-master branch
  46. pcmattmangit diff in the main Pedigree git...
  47. pcmattman-Subproject commit 2474abcfb6d451e15360725e88b43aa0a0c19814
  48. pcmattman+Subproject commit 4db03c5ebd62e58af4d72afaf0e080faa89a6f41
  49. pcmattmanNo mention of branch, just a hash
  50. kwolfSo it just works
  51. pcmattmanIf a hash is not linked to any branch, yes :)
  52. kwolfIf I understand it right, a branch is technically just a pointer to the head of that branch
  53. pcmattmanRight
  54. pcmattmanIt just follows the family tree of the commit to find the branch's history, I guess
  55. kwolfSame as with the history for a commit which isn't a head
  56. pcmattmanWell, the magic of git ends up getting figured out eventually when you start doing crazy stuff like subprojects :)
  57. kwolfSubmodules look pretty harmless, actually
  58. pcmattmanI think though to create a branch which I can push into can really trash the workflow
  59. pcmattmanAnd yeah, once you figure out some of the little "fun details" with subprojects they end up being pretty sane
  60. kwolfHow would it trash the workflow?
  61. pcmattmanWell, at the moment anything I commit is reviewed first
  62. pcmattmanThat would still happen for the CDI master, but it could cause some fun issues with the state of the branch compared to master
  63. kwolfWell, either this happens to be an advantage, then there's no reason to stop it
  64. kwolfOr you run into serious quality issues, then you should think about the way you use your branch ;-)
  65. kwolfI would expect your branch to rebase on master regularly and drop patches that have been changed after review, but that's up to you
  66. kwolfBasically, if there is something that breaks the Pedigree build, committing into a branch would help immediately. You could leave the branch completely unused when master is working.
  67. pcmattmanUsually the issue is resolved in good time anyway
  68. pcmattmanI'm just not sure having an alternative branch with full push access is the right answer. But we'll discuss it further, see the pros and cons...
  69. pcmattmanI'm out for a few hours now though, so I'll be back later
  70. kwolfWell, it's just an offer. If you don't want it, I won't insist on it
  71. kwolfOk
  72. eddybmoin
  73. XanClicMoin, eddyb
  74. eddybwhere's my magnifying penguin?
  75. eddybXanClic: what
  76. eddyb*'s data toggle for?
  77. XanClicI know that at least for bulk transfers
  78. XanClicHm, but it's also used on control...
  79. XanClicYou send DATA0 first, then DATA1, then DATA0, ...
  80. eddybso i *must* use it?
  81. XanClicYes
  82. XanClicEspecially on control transfers
  83. XanClicThe last packet you send there is always DATA1, afair
  84. eddyband for same combination of address and endpoint i have to keep track of last packet?
  85. XanClicRead the osdev wiki page on USB, it'll explain that ;-)
  86. eddyb*'s data toggle
  87. XanClicYes, on toggle per pipe
  88. XanClic*one
  89. eddybmeh, i like human explanations more :P
  90. eddybthis is going to be a PITA
  91. XanClicUSB is ugly
  92. eddybhmm, it works in qemu without touching data toggle
  93. XanClicBut it won't work on real hardware.
  94. eddybin EHCI there's a bit in QH called data toggle control
  95. XanClicYou're forced to use that data toggle thing
  96. XanClicUHCI also has such a bit
  97. XanClicIt's a bit simpler with OHCI again ;-)
  98. eddybUHCI got a data toggle
  99. eddybin TD
  100. XanClicBecause the OHCD tracks all your pipes
  101. eddybqTD in EHCI, too
  102. XanClicAnd it also takes care of setting the data toggle bit appropriately
  103. eddybi mean
  104. eddybyou can tell EHCI to not use your data toggle
  105. eddyband maybe it does it by itself
  106. XanClicHm, but I doubt if it handles control transfers correctly
  107. eddybbtw
  108. XanClicDoes an EHC handle pipes?
  109. XanClicI mean, you define a USB address and an endpoint and then you can send packets there
  110. eddybwhy do i have to do a IN with no size in order to "clear" a SETUP that doesn't receive or send anything?
  111. XanClicIt was the handshake, wasn't it?
  112. eddybwhatever
  113. XanClicJust read the wiki page
  114. XanClicIt's really good
  115. eddyband you really have to dig up information
  116. eddybbut it helped me with control stuff
  117. XanClicYes, if you don't want to read all of it...
  118. XanClicHm, let's see when kde 4.4 is available...
  119. XanClicAh, not yet.
  120. eddybdo you have kde nightly?
  121. eddybor how is called
  122. XanClicNo, just in the standard openSUSE repos
  123. XanClicI know the source is available and so on
  124. eddybUNSTABLE....
  125. XanClicBah ;-)
  126. eddybsomething from the last update
  127. eddybRetrieving: oxygen-icon-theme-large-4.4.63svn1087770-1.1.noarch.rpm [done (168,5 KiB/s)]
  128. eddybah
  129. eddybthat thing in EHCI isn't what i was thinking....
  130. eddyb:(
  131. eddyb128*16....
  132. eddybit'd take 256 bytes with a bitset
  133. eddybnot so much
  134. eddyblol, i got 16 bytes to play with :)
  135. eddybvery useful with a dumb controller
  136. eddybaaah
  137. eddybcerr
  138. eddybXanClic: can it be that HC retries to infinite to send mah packet?
  139. XanClicI think not
  140. eddybwhy?
  141. XanClicIt should signal an IRQ and let you decide what to do
  142. eddybi had cerr = 0
  143. XanClicAt least UHCI should do that
  144. XanClicThere are error conditions such as NAK, aren't there?
  145. XanClicAnd NAK is the only case where it could retry
  146. eddybBits[28:27]
  147. eddyb00
  148. XanClicThe other ones are more ore less "serious" errors
  149. eddyb01
  150. eddybInterrupt After
  151. eddybNo Error Limit
  152. eddyb1 error
  153. eddybargh
  154. eddybokteta copies wrong
  155. eddybanyway
  156. eddybInterrupt after... No error limit :)
  157. XanClicSo you should set an error limit ;-)
  158. eddyb1, to be exactly
  159. eddybi changed my TD structure to bit(whatever) to be more like EHCI and i saw cerr
  160. eddybif it works now i killed two rabbits from one shot...
  161. eddybit still doesn't explain what's happening to EHCI
  162. eddybfunny
  163. eddybserial is used
  164. eddybstill, i got serial output only when an irq comes in
  165. eddybah, whoops
  166. eddybi left EHCI XD
  167. eddyb(NN) 0x284
  168. eddyb(NN) 0x285
  169. eddybwtf?
  170. eddybi finally get connected bit set :O ?
  171. eddybstill, no success :(
  172. eddybone way to try....
  173. eddybdo something nasty...
  174. eddybi added a +1 after pid :P
  175. eddybwriting to reserved fields, nope....
  176. eddybnow i see i got two messages
  177. eddybPM
  178. eddybDelete message
  179. eddybThis message has been removed by its author before it was delivered.
  180. eddybby neato on Thu Feb 11, 2010 11:26 pm
  181. eddybwtf?
  182. eddybso i can't see that one
  183. pcmattmanBack for a little while
  184. eddybbut the other has a quote to my whole last post and under it "I'm sorry little girl, don't cry."
  185. eddyb?!
  186. eddybpcmattman: sorry for the backlog :D
  187. pcmattmanNot my channel
  188. eddybi mean if you want to read it :P
  189. eddybanyway, it's all my usb shit
  190. pcmattmanI don't want to read it
  191. kwolfHeh, right, as long as we get a USB driver that we can just use ;-)
  192. eddybcan someone tell me what the n00b wanted to tell me?
  193. eddybargh
  194. pcmattmanI'll just go grab my crystal ball
  195. eddybpcmattman: 0x1337cafe strikes again
  196. eddybmaybe you guys know better
  197. pcmattmanClosest I can do is guess what he wanted to say
  198. pcmattmanYou don't want to hear it
  199. pcmattmanAnd as for 0x1337cafe, you do realise these bugs only happen when your code is wrong
  200. pcmattmanIt's not Pedigree, it's your code