Activity for John Huff
-
John Huff posted a comment on ticket #189
I actually don't have a use case for changing heapmax during runtime. I've been developing an HTTP API for semi-remote (i.e. using terrestrial Internet) ION node management. I just happened to stumble across this as I was testing my API.
-
Note that ltpdb->maxAcqInHeap is used elsewhere for allocating buffers such as in ltpdeliv. So if you increase the size of any session buffers while ltpdeliv is still running, the ltpdeliv buffer will overflow. This should also be addressed. Something simple you could do would be to define a maximum allowed value for maxAcqInHeap such as 64KB, and allocate a buffer of that size in ltpdeliv. This is ineffcient but ensures that ltpdeliv always has a big enough buffer.
-
LTP heap buffer overflow
-
Primary block is mutated due to incomplete CRC support
-
Actually, the handleAdminBundles function does correctly handle the SIGTERM signal, ending the BpSAP recvSemaphore. However, handleAdminBundles doesn't close the BpSAP on exit.
-
scheme adminep daemons leak psm memory on stop
-
this was from me btw, forgot to log in
-
Yes, my understanding is that it should be newRoute. newRoute was correctly having toNodeNbr and fromTime set in computeRoute, making this bug harder to detect. Computing a spur route shouldn't have any effect on the route used to compute it.
-
Unable to remove LTP span once started
-
Missing support for large, fragmented bundles
-
title "should be ipnfw ipn_lookupOvrd arguments reversed", don't know if you can edit tickets though
-
ipnfw ipn_lookupOvrd arguments revered
-
'm home' and 'm outer' ionadmin commands
-
edit: sm_Unend -> sm_SemUnend
-
LTP deliverySemaphore is never unended
-
Fixed crash caused by llcv_closed.
-
Fixed issue of tcpcli not shutting down
-
Fixed issue causeing bug-0015 test to fail. Checks if llcv is closed before signalling, prevents failed assertion.
-
Added changes to fix tcpcli shutdown, more narrowly this time.
-
Slight changes to llcv_close to fix tcp issues.
-
Merged changes
-
Increased sleep to allow file to finish sending
-
Merge with tcp-fix-3.6.2
-
Reverted reordering of mutex unlock and cond destory, and instead put another mutex_unlock if cond_destroy fails.
-
Reordered mutex unlock and cond destory.
-
BSD uses pthread_set_name_np rather than pthread_setname_np
-
Added thread naming. Overloaded pthread_begin to have an extra parameter
-
Named all threads
-
Added configure error when setting BUILD_32_ON_64_SOLARS=1 and not setting CC={solaris studio cc}
-
Added flag BUILD_32_ON_64_SOLARIS.
-
Added thread naming on OSX by wrapping pthread_begin
-
Fix for Solaris 64-bit BpTimestamp bug. time_t can be 8 bytes on some 64-bit systems, and a 32-bit value was being extracted into the first 4 bytes of an 8 byte time_t because it was being referenced to by a 4 byte unsigned int pointer. This caused expiredTTL events because the time_t variable was later casted to unsigned int which always had a value of zero. Fix is to first extract into a unsigned int and then cast to a time_t.
-
Added --enable-high-speed option for configure. Increases
-
Fixed bug with reception rate limiting. Receiver was not taking into account the amount of time it takes to read the data. Code now resembles the throttle mechanism that is used on the sender. Rate limiting already occurs on the sender, so it may be better to altogether remove the rate limiting on the receiver.
-
Added CRC-32 support (alongside the already supported CRC-32C).
-
fixed redefine in crc32_16bytes
-
Merged with ion-4.0.0-high-speed. Removed rate limiting on TCPCLA receiver. This is not needed as the sender handles rate limiting and it uses TCP which has its own sort of rate limiting
-
removed leftover MIN_SNOOZE macro
1
